Models · OpenAI · Out sinceReleased 9 Jul 2026

GPT-5.6 Sol#9 for writing.#9 for writing, best at extra high effort.

Available on OpenRouterOn OpenRouterAvailable nowGenerally availableClosed (can't be downloaded)Proprietary

Where you can use it:In apps:ChatGPT (Business plan) · Instant / Medium / High / Extra High

GPT-5.6 Sol is made by OpenAI. Among the models we track it ranks #9 for writing, #11 for coding, #14 for research and analysis. It's expensive to use.

Launched at $5/$30; promotional $4/$20 since Aug 21, 2026 (through at least Nov 21, 2026) per OpenAI docs; OpenRouter lists $2/$10. Alias gpt-5.6. Default effort medium. "ultra" = multi-agent mode (4 parallel agents) in ChatGPT/Codex. Preview Jun 26, GA Jul 9, 2026.

Writing & creativity: 61.3 / 100 · #9
Research & analysis: 71.1 / 100 · #14
Coding: 58.3 / 100 · #11
Price: Expensive$4 / $20
Price per 1M (blended)Blended / 1M: $8
MemoryContext: 1.1M
Longest answerMax output: 128K
Test resultsResults: 217 (136 independent136 indep.)
Out sinceReleased: 9 Jul 2026
Made byVendor: OpenAI
UnderstandsInputs: text, image

Scores are out of 100 for each area, compared with every model we track; “#” is its rank. Context is how much text it can read at once (1M is roughly 700,000 words). Blended price mixes the cost of what you send and what it writes back.

Thinking levels

Reasoning settings

No reasoningLowMediumHighExtra highMax

Highlighted: where it did best for writing (Extra high thinking).

Highlighted: dominant setting in its writing composite.

Links

Using it through OpenRouter

On OpenRouter

OpenRouter is a service that gives access to many AI models in one place. This model is listed there as openai/gpt-5.6-sol.

Route it as openai/gpt-5.6-sol at $2 in / $10 out per 1M tokens, 1.1M context. Listed since 9 Jul 2026.

include_reasoningmax_completion_tokensmax_tokensreasoningreasoning_effortresponse_formatseedstructured_outputstool_choicetools

Thinking level

Reasoning effort

How long should GPT-5.6 SolWhere GPT-5.6 Solthink?peaks.

Each chart is one test. Left to right, the model thinks longer; higher is a better result. The yellow ring marks where it did best. Thinking level is how long the AI thinks before it answers. Higher is slower and costs more, and it isn't always better.

Score per reasoning setting on each benchmark with ≥2 settings; ring = peak. Canonical results (independent over vendor).

Artificial Analysis Coding Agent Index

Best at High, worse abovePeaks at High

Terminal-Bench 4.0

Best at MaxBest at Max

CursorBench

Best at MaxBest at Max

DeepSWE

Best at MaxBest at Max

FrontierCode

Best at MaxBest at Max

SWE Atlas - Codebase QnA

Best at MaxBest at Max

ProgramBench (fully resolved)

Best at MaxBest at Max

Terminal-Bench 2.1

Best at Extra high, worse abovePeaks at Extra high

SciCode

Best at High, worse abovePeaks at High

Agents' Last Exam

Best at Extra high, worse abovePeaks at Extra high

Artificial Analysis Intelligence Index

Best at MaxBest at Max

AutomationBench

Best at MaxBest at Max

OSWorld 2.0 offline set

Best at MaxBest at Max

tau2-bench

Best at MaxBest at Max

ARC-AGI-2

Best at MaxBest at Max

ARC-AGI-3

Best at MaxBest at Max

GDPval-AA v2.1

Best at MaxBest at Max

GPQA Diamond

Best at MaxBest at Max

Humanity's Last Exam

Best at MaxBest at Max

LLM Creative Story-Writing Benchmark (Lech Mazur)

Best at Extra highBest at Extra high

ScreenSpot-Pro

Best at Extra high, worse abovePeaks at Extra high

Every test result

Every result

All the numbers,with where they came from.

Every result we've found for this model, with who measured it and a link to where we read it.

All raw rows (vendor and independent kept separate), with setting label, source, date, harness and notes.

All results for this model
						NotesNotes
AA Analyst Agent	47.5%	MaxGPT-5.6 Sol (Max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run; scores move in 1.25-pt steps (small task set)
Agents' Last Exam	45.1%	Lowreasoning effort=low	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	Agents' Last Exam V1 (long-horizon professional workflows, 55 sub-industries); vendor-estimated API cost/task $1.6324
Agents' Last Exam	52.1%	Mediumreasoning effort=medium	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	Agents' Last Exam V1 (long-horizon professional workflows, 55 sub-industries); vendor-estimated API cost/task $3.4126
Agents' Last Exam	52.4%	Highreasoning effort=high	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	Agents' Last Exam V1 (long-horizon professional workflows, 55 sub-industries); vendor-estimated API cost/task $3.7874
Agents' Last Exam	53.6%	Extra highreasoning effort=xhigh	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	Agents' Last Exam V1 (long-horizon professional workflows, 55 sub-industries); vendor-estimated API cost/task $5.0764
Agents' Last Exam	52.8%	Maxreasoning effort=max	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	Agents' Last Exam V1 (long-horizon professional workflows, 55 sub-industries); vendor-estimated API cost/task $7.1322
Agents' Last Exam	52.7%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
ARC-AGI-1	97.5%	Defaultbest score across efforts (OpenAI table: "maximum at any effort")	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—
ARC-AGI-2	85.4%	HighGPT-5.6 Sol (High)	Independent testIndependentARC Prize ↗	30 Sep 2026	—	semi-private set; $0.74/task; data https://arcprize.org/media/data/leaderboard/v2.json
ARC-AGI-2	90.0%	Extra highGPT-5.6 Sol (XHigh)	Independent testIndependentARC Prize ↗	30 Sep 2026	—	semi-private set; $1.04/task; data https://arcprize.org/media/data/leaderboard/v2.json
ARC-AGI-2	92.5%	MaxGPT-5.6 Sol (Max)	Independent testIndependentARC Prize ↗	30 Sep 2026	—	semi-private set; $1.44/task; data https://arcprize.org/media/data/leaderboard/v2.json
ARC-AGI-2	92.5%	Defaultbest score across efforts (OpenAI table: "maximum at any effort")	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—
ARC-AGI-3	0.3%	LowGPT-5.6 Sol (Low)	Independent testIndependentARC Prize ↗	30 Sep 2026	—	semi-private set; total cost $12,806; data https://arcprize.org/media/data/leaderboard/v3.json
ARC-AGI-3	1.1%	MediumGPT-5.6 Sol (Medium)	Independent testIndependentARC Prize ↗	30 Sep 2026	—	semi-private set; total cost $12,971; data https://arcprize.org/media/data/leaderboard/v3.json
ARC-AGI-3	2.1%	HighGPT-5.6 Sol (High)	Independent testIndependentARC Prize ↗	30 Sep 2026	—	semi-private set; total cost $15,176; data https://arcprize.org/media/data/leaderboard/v3.json
ARC-AGI-3	7.0%	Extra highGPT-5.6 Sol (XHigh)	Independent testIndependentARC Prize ↗	30 Sep 2026	—	semi-private set; total cost $19,216; data https://arcprize.org/media/data/leaderboard/v3.json
ARC-AGI-3	7.8%	MaxGPT-5.6 Sol (Max)	Independent testIndependentARC Prize ↗	30 Sep 2026	—	semi-private set; total cost $25,064; data https://arcprize.org/media/data/leaderboard/v3.json
ARC-AGI-3	7.8%	Defaultbest score across efforts (OpenAI table: "maximum at any effort")	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—
ARC-AGI-3	7.8%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
Artificial Analysis Coding Agent Index	43.4	No reasoningreasoning effort=none	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	Artificial Analysis Coding Agent Index v1.4 (index score 0-100); vendor-estimated API cost/task $1.09
Artificial Analysis Coding Agent Index	55.2	Lowreasoning effort=low	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	Artificial Analysis Coding Agent Index v1.4 (index score 0-100); vendor-estimated API cost/task $1.29
Artificial Analysis Coding Agent Index	61.6	Mediumreasoning effort=medium	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	Artificial Analysis Coding Agent Index v1.4 (index score 0-100); vendor-estimated API cost/task $2.19
Artificial Analysis Coding Agent Index	64.1	Highreasoning effort=high	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	Artificial Analysis Coding Agent Index v1.4 (index score 0-100); vendor-estimated API cost/task $3
Artificial Analysis Coding Agent Index	63.3	Extra highreasoning effort=xhigh	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	Artificial Analysis Coding Agent Index v1.4 (index score 0-100); vendor-estimated API cost/task $3.74
Artificial Analysis Coding Agent Index	54.6	MaxGPT-5.6 Sol (max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	Codex	agent Codex; components: DeepSWE v1.1 72.3, SWE-Atlas-QnA 54.0, Terminal-Bench v4 37.4; avg cost $6.35/task; avg wall time 21 min/task
Artificial Analysis Coding Agent Index	65.1	Maxreasoning effort=max	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	Artificial Analysis Coding Agent Index v1.4 (index score 0-100); vendor-estimated API cost/task $5
Artificial Analysis Coding Agent Index v1.1	80	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—	Artificial Analysis Coding Agent Index v1.1 (index score)
Artificial Analysis Intelligence Index	33.5	LowGPT-5.6 Sol (Low)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug gpt-5-6-sol-low; list price $4/20 per 1M in/out; cost to run AA Intelligence Index $0.26/task (AA marks this variant deprecated)
Artificial Analysis Intelligence Index	39.2	MediumGPT-5.6 Sol (Medium)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug gpt-5-6-sol-medium; list price $4/20 per 1M in/out; cost to run AA Intelligence Index $0.50/task (AA marks this variant deprecated)
Artificial Analysis Intelligence Index	42.3	HighGPT-5.6 Sol (High)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug gpt-5-6-sol-high; list price $4/20 per 1M in/out; cost to run AA Intelligence Index $0.81/task (AA marks this variant deprecated)
Artificial Analysis Intelligence Index	44	Extra highGPT-5.6 Sol (Xhigh)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug gpt-5-6-sol-xhigh; list price $4/20 per 1M in/out; cost to run AA Intelligence Index $1.18/task (AA marks this variant deprecated)
Artificial Analysis Intelligence Index	47	MaxGPT-5.6 Sol (Max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug gpt-5-6-sol; list price $4/20 per 1M in/out; cost to run AA Intelligence Index $1.99/task (AA marks this variant deprecated)
Artificial Analysis Intelligence Index	60.9	Defaultbest score across efforts (OpenAI table: "maximum at any effort")	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	v4.1.1
Artificial Analysis Intelligence Index	58.9	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—	Artificial Analysis Intelligence Index v4.1
Artificial Analysis output speed	67 tok/s	LowGPT-5.6 Sol (Low)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	tokens/sec (median output speed, first-party API); TTFT 2.4s; list price $4/20 per 1M in/out (AA marks this variant deprecated)
Artificial Analysis output speed	70 tok/s	MediumGPT-5.6 Sol (Medium)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	tokens/sec (median output speed, first-party API); TTFT 6.7s; list price $4/20 per 1M in/out (AA marks this variant deprecated)
Artificial Analysis output speed	74 tok/s	HighGPT-5.6 Sol (High)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	tokens/sec (median output speed, first-party API); TTFT 16.4s; list price $4/20 per 1M in/out (AA marks this variant deprecated)
Artificial Analysis output speed	76 tok/s	Extra highGPT-5.6 Sol (Xhigh)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	tokens/sec (median output speed, first-party API); TTFT 53.9s; list price $4/20 per 1M in/out (AA marks this variant deprecated)
Artificial Analysis output speed	79 tok/s	MaxGPT-5.6 Sol (Max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	tokens/sec (median output speed, first-party API); TTFT 119.5s; list price $4/20 per 1M in/out (AA marks this variant deprecated)
AutomationBench	11.7%	Lowreasoning effort=low	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	AutomationBench 1.0.6 (Zapier); vendor-estimated API cost/task $0.31
AutomationBench	19.6%	Mediumreasoning effort=medium	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	AutomationBench 1.0.6 (Zapier); vendor-estimated API cost/task $0.42
AutomationBench	24.8%	Highreasoning effort=high	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	AutomationBench 1.0.6 (Zapier); vendor-estimated API cost/task $0.47
AutomationBench	26.3%	Extra highreasoning effort=xhigh	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	AutomationBench 1.0.6 (Zapier); vendor-estimated API cost/task $0.54
AutomationBench	28.8%	Maxreasoning effort=max	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	AutomationBench 1.0.6 (Zapier); vendor-estimated API cost/task $0.67
AutomationBench	18.1%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
BrowseComp	92.2%	DefaultGPT-5.6 Sol Ultra (4 parallel agents)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—	ultra = multi-agent parallel test-time compute
BrowseComp	90.4%	Defaultbest score across efforts (OpenAI table: "maximum at any effort")	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—
BrowseComp	90.4%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
CursorBench	24.6%	LowGPT-5.6 Sol Low	Independent testIndependentCursor (CursorBench) ↗	1 Oct 2026	Cursor agent	CursorBench 4.0 (tasks from real Cursor sessions; v4.0 introduced 2026-09-10); rank 59; $0.87/task; 4,885 tokens/task; 21 steps/task
CursorBench	24.6%	Lowreasoning effort=low	Independent testIndependentCursor (via Anthropic launch post) ↗	22 Sep 2026	Cursor agent	CursorBench 4.0, run end-to-end by Cursor in its production agent harness; cost/task $0.87 as published by Cursor; OpenAI has not published CursorBench 4.0 for GPT-6 models
CursorBench	31.1%	MediumGPT-5.6 Sol Medium	Independent testIndependentCursor (CursorBench) ↗	1 Oct 2026	Cursor agent	CursorBench 4.0 (tasks from real Cursor sessions; v4.0 introduced 2026-09-10); rank 48; $1.77/task; 10,111 tokens/task; 32 steps/task
CursorBench	31.1%	Mediumreasoning effort=med	Independent testIndependentCursor (via Anthropic launch post) ↗	22 Sep 2026	Cursor agent	CursorBench 4.0, run end-to-end by Cursor in its production agent harness; cost/task $1.77 as published by Cursor; OpenAI has not published CursorBench 4.0 for GPT-6 models
CursorBench	35.7%	HighGPT-5.6 Sol High	Independent testIndependentCursor (CursorBench) ↗	1 Oct 2026	Cursor agent	CursorBench 4.0 (tasks from real Cursor sessions; v4.0 introduced 2026-09-10); rank 38; $2.85/task; 16,174 tokens/task; 41 steps/task
CursorBench	35.7%	Highreasoning effort=high	Independent testIndependentCursor (via Anthropic launch post) ↗	22 Sep 2026	Cursor agent	CursorBench 4.0, run end-to-end by Cursor in its production agent harness; cost/task $2.85 as published by Cursor; OpenAI has not published CursorBench 4.0 for GPT-6 models
CursorBench	37.7%	Extra highGPT-5.6 Sol Extra High	Independent testIndependentCursor (CursorBench) ↗	1 Oct 2026	Cursor agent	CursorBench 4.0 (tasks from real Cursor sessions; v4.0 introduced 2026-09-10); rank 31; $4.40/task; 24,729 tokens/task; 55 steps/task
CursorBench	37.7%	Extra highreasoning effort=xhigh	Independent testIndependentCursor (via Anthropic launch post) ↗	22 Sep 2026	Cursor agent	CursorBench 4.0, run end-to-end by Cursor in its production agent harness; cost/task $4.4 as published by Cursor; OpenAI has not published CursorBench 4.0 for GPT-6 models
CursorBench	41.7%	MaxGPT-5.6 Sol Max	Independent testIndependentCursor (CursorBench) ↗	1 Oct 2026	Cursor agent	CursorBench 4.0 (tasks from real Cursor sessions; v4.0 introduced 2026-09-10); rank 21; $8.23/task; 42,944 tokens/task; 99 steps/task
CursorBench	41.7%	Maxreasoning effort=max	Independent testIndependentCursor (via Anthropic launch post) ↗	22 Sep 2026	Cursor agent	CursorBench 4.0, run end-to-end by Cursor in its production agent harness; cost/task $8.23 as published by Cursor; OpenAI has not published CursorBench 4.0 for GPT-6 models
CursorBench 3.2.0	67.2%	Maxreasoning effort=max	Independent testIndependentCursor (via Anthropic launch post) ↗	1 Sep 2026	Cursor agent	CursorBench 3.2.0 (Cursor production harness); not comparable to CursorBench 4.0; cost $5.69/task
DeepSWE	45.4%	Lowreasoning effort=low	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	Codex	DeepSWE v1.1 (113 long-horizon SWE tasks); GPT models run in Codex; vendor-estimated API cost/task $0.82
DeepSWE	61.1%	Mediumgpt-5.6-sol_medium	Independent testIndependentDeepSWE (Datacurve) via Epoch AI ↗	1 Oct 2026	mini-SWE-agent	read from Epoch AI benchmark_data.zip (deepswe_external.csv); original leaderboard https://deepswe.datacurve.ai/; harness mini-swe-agent; pass@4 80.5%; ±1.6; 4 runs; $1.86/task
DeepSWE	61.1%	Mediumreasoning effort=medium	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	Codex	DeepSWE v1.1 (113 long-horizon SWE tasks); GPT models run in Codex; vendor-estimated API cost/task $1.42
DeepSWE	69.4%	Highgpt-5.6-sol_high	Independent testIndependentDeepSWE (Datacurve) via Epoch AI ↗	1 Oct 2026	mini-SWE-agent	read from Epoch AI benchmark_data.zip (deepswe_external.csv); original leaderboard https://deepswe.datacurve.ai/; harness mini-swe-agent; pass@4 86.7%; ±1.4; 4 runs; $3.47/task
DeepSWE	69.4%	Highreasoning effort=high	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	Codex	DeepSWE v1.1 (113 long-horizon SWE tasks); GPT models run in Codex; vendor-estimated API cost/task $2.66
DeepSWE	70.7%	Extra highgpt-5.6-sol_xhigh	Independent testIndependentDeepSWE (Datacurve) via Epoch AI ↗	1 Oct 2026	mini-SWE-agent	read from Epoch AI benchmark_data.zip (deepswe_external.csv); original leaderboard https://deepswe.datacurve.ai/; harness mini-swe-agent; pass@4 85.8%; ±0.8; 4 runs; $4.70/task
DeepSWE	70.7%	Extra highreasoning effort=xhigh	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	Codex	DeepSWE v1.1 (113 long-horizon SWE tasks); GPT models run in Codex; vendor-estimated API cost/task $3.6
DeepSWE	72.7%	Maxgpt-5.6-sol_max	Independent testIndependentDeepSWE (Datacurve) via Epoch AI ↗	1 Oct 2026	mini-SWE-agent	read from Epoch AI benchmark_data.zip (deepswe_external.csv); original leaderboard https://deepswe.datacurve.ai/; harness mini-swe-agent; pass@4 85.8%; ±2.8; 4 runs; $8.39/task
DeepSWE	72.3%	MaxGPT-5.6 Sol (max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	Codex	AA Coding Agent Index component; pass@1 avg of 3 attempts
DeepSWE	72.7%	Maxreasoning effort=max	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	Codex	DeepSWE v1.1 (113 long-horizon SWE tasks); GPT models run in Codex; vendor-estimated API cost/task $6.46
DeepSWE	72.7%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	Codex
Design Arena (all categories)	1345	Extra highgpt-5.6-sol-xhigh	Independent testIndependentDesign Arena ↗	1 Oct 2026	—	Bradley-Terry Elo; ±3.7 SE; 9719 battles; win rate 58.6%
Design Arena (all categories)	1325	Defaultgpt-5.6-sol	Independent testIndependentDesign Arena ↗	1 Oct 2026	—	Bradley-Terry Elo; ±2.5 SE; 24080 battles; win rate 57.4%
Design Arena (fullstack)	1259	Extra highgpt-5.6-sol-xhigh	Independent testIndependentDesign Arena ↗	1 Oct 2026	—	Bradley-Terry Elo; ±8.5 SE; 1973 battles; win rate 54.3%
Design Arena (fullstack)	1157	Defaultgpt-5.6-sol	Independent testIndependentDesign Arena ↗	1 Oct 2026	—	Bradley-Terry Elo; ±9.2 SE; 1614 battles; win rate 49.8%
Epoch Capabilities Index	161.8	Default	Independent testIndependentEpoch AI ↗	1 Oct 2026	—	Epoch Capabilities Index; 90% CI 159.3-165.3; best of listed model versions
EQ-Bench Creative Writing v3 (Elo)	1972	Defaultgpt-5.6-sol	Independent testIndependentEQ-Bench ↗	1 Oct 2026	—	leaderboard rank 10; rubric score 16.78/20; slop 11.68; avg length 8548 chars; Elo judged by Claude Sonnet 4.6; effort not stated by EQ-Bench
EuroEval Swedish (generative)	1.3	Defaultgpt-5.6-sol (zero-shot, val)	Independent testIndependentEuroEval (Alexandra Institute) ↗	29 Sep 2026	—	EuroEval rank tier 2; ±0.04; lower is better; task scores (first metric): SweDN summarisation 37.92 ± 0.22, Skolprov 90.97 ± 1.67, Swedish facts 85.67 ± 1.96, ScaLA-sv 72.02 ± 1.52
FrontierCode	35.4%	Lowreasoning effort=low	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	Codex	FrontierCode 1.1 Main (Cognition); GPT models run in Codex CLI; vendor-estimated API cost/task $1.89
FrontierCode	39.9%	Mediumreasoning effort=medium	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	Codex	FrontierCode 1.1 Main (Cognition); GPT models run in Codex CLI; vendor-estimated API cost/task $2.69
FrontierCode	45.1%	Highreasoning effort=high	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	Codex	FrontierCode 1.1 Main (Cognition); GPT models run in Codex CLI; vendor-estimated API cost/task $3.48
FrontierCode	46.8%	Extra highreasoning effort=xhigh	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	Codex	FrontierCode 1.1 Main (Cognition); GPT models run in Codex CLI; vendor-estimated API cost/task $4.15
FrontierCode	47.5%	Maxreasoning effort=max	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	Codex	FrontierCode 1.1 Main (Cognition); GPT models run in Codex CLI; vendor-estimated API cost/task $5.19
FrontierCode	47.5%	Defaultgpt-5.6-sol_unknown	Independent testIndependentFrontierCode (Cognition) via Epoch AI ↗	1 Oct 2026	codex	read from Epoch AI benchmark_data.zip (frontiercode_external.csv); original leaderboard https://cognition.com/frontiercode; harness codex; Mean@5
FrontierCode v1.1 (Extended)	60.6%	Defaultbest score across efforts (OpenAI table: "maximum at any effort")	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—
FrontierMath (Tiers 1-3)	89.1%	Maxgpt-5.6-sol_max	Independent testIndependentEpoch AI ↗	9 Jul 2026	—	FrontierMath Tiers 1-3 (v2), Epoch-run; stderr 1.8pt; data https://epoch.ai/data/benchmark_data.zip (frontiermath_tiers_1_3_v2.csv)
FrontierMath Tier 1-3	89.0%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
FrontierMath Tier 4	82.9%	Maxgpt-5.6-sol_max	Independent testIndependentEpoch AI ↗	9 Jul 2026	—	FrontierMath Tier 4 (v2), Epoch-run; stderr 5.9pt; data https://epoch.ai/data/benchmark_data.zip (frontiermath_tier_4_v2.csv)
FrontierMath Tier 4	83.0%	Defaultbest score across efforts (OpenAI table: "maximum at any effort")	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	v2
FrontierMath Tier 4	83.0%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
FrontierSWE v2	32.2%	Maxreasoning effort=max	Independent testIndependentProximal (via Anthropic system card) ↗	22 Sep 2026	Proximal harness
GDP.pdf	30.7%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
GDPval-AA v2	1711	Defaultas listed by Anthropic	Independent testIndependentArtificial Analysis (via Anthropic launch post) ↗	1 Sep 2026	—
GDPval-AA v2	1748	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
GDPval-AA v2.1	1289	Lowreasoning effort=low	Independent testIndependentArtificial Analysis (via Anthropic launch post) ↗	22 Sep 2026	—	GDPval-AA v2.1 (Artificial Analysis, Elo from blind pairwise comparisons, 44 occupations); cost/task $0.27
GDPval-AA v2.1	1403	Mediumreasoning effort=med	Independent testIndependentArtificial Analysis (via Anthropic launch post) ↗	22 Sep 2026	—	GDPval-AA v2.1 (Artificial Analysis, Elo from blind pairwise comparisons, 44 occupations); cost/task $0.6
GDPval-AA v2.1	1480	Highreasoning effort=high	Independent testIndependentArtificial Analysis (via Anthropic launch post) ↗	22 Sep 2026	—	GDPval-AA v2.1 (Artificial Analysis, Elo from blind pairwise comparisons, 44 occupations); cost/task $1.11
GDPval-AA v2.1	1548	Extra highreasoning effort=xhigh	Independent testIndependentArtificial Analysis (via Anthropic launch post) ↗	22 Sep 2026	—	GDPval-AA v2.1 (Artificial Analysis, Elo from blind pairwise comparisons, 44 occupations); cost/task $1.7
GDPval-AA v2.1	1588	Maxreasoning effort=max	Independent testIndependentArtificial Analysis (via Anthropic launch post) ↗	22 Sep 2026	—	GDPval-AA v2.1 (Artificial Analysis, Elo from blind pairwise comparisons, 44 occupations); cost/task $2.81
GPQA Diamond	89.8%	LowGPT-5.6 Sol (Low)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-low) (AA marks this variant deprecated)
GPQA Diamond	91.7%	Lowreasoning effort=low	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	GPQA Diamond; vendor-estimated API cost/task $0.0197313
GPQA Diamond	92.6%	MediumGPT-5.6 Sol (Medium)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-medium) (AA marks this variant deprecated)
GPQA Diamond	92.9%	Mediumreasoning effort=medium	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	GPQA Diamond; vendor-estimated API cost/task $0.0298166
GPQA Diamond	92.8%	HighGPT-5.6 Sol (High)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-high) (AA marks this variant deprecated)
GPQA Diamond	93.2%	Highreasoning effort=high	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	GPQA Diamond; vendor-estimated API cost/task $0.048927
GPQA Diamond	93.1%	Extra highGPT-5.6 Sol (Xhigh)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-xhigh) (AA marks this variant deprecated)
GPQA Diamond	93.9%	Extra highreasoning effort=xhigh	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	GPQA Diamond; vendor-estimated API cost/task $0.0687297
GPQA Diamond	95.2%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	1 Sep 2026	—	±1.074 stderr; $0.110/test
GPQA Diamond	94.1%	MaxGPT-5.6 Sol (Max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol) (AA marks this variant deprecated)
GPQA Diamond	93.5%	Maxgpt-5.6-sol_max	Independent testIndependentEpoch AI ↗	9 Jul 2026	—	Epoch-run GPQA Diamond; stderr 1.6pt; data https://epoch.ai/data/benchmark_data.zip (gpqa_diamond.csv)
GPQA Diamond	94.6%	Maxreasoning effort=max	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	GPQA Diamond; vendor-estimated API cost/task $0.1267
GPQA Diamond	94.6%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
GSO	76.5%	Extra highreasoning_effort=xhigh	Independent testIndependentGSO ↗	27 Sep 2026	OpenHands	Opt@1; hack-controlled score 70.59; data https://gso-bench.github.io/assets/leaderboard.json
HealthBench Professional	60.5%	Defaultbest score across efforts (OpenAI table: "maximum at any effort")	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	length-adjusted
HealthBench Professional	60.5%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—	official HealthBench Professional scoring
HLE Diamond	31.2%	Defaultgpt-5.6-sol	Independent testIndependentScale AI SEAL ↗	1 Oct 2026	—	rank 7 (Scale rank accounts for CI); ±2.9; entry added 2025-11-19
Humanity's Last Exam	39.4%	LowGPT-5.6 Sol (Low)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-low) (AA marks this variant deprecated)
Humanity's Last Exam	42.2%	MediumGPT-5.6 Sol (Medium)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-medium) (AA marks this variant deprecated)
Humanity's Last Exam	46.0%	HighGPT-5.6 Sol (High)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-high) (AA marks this variant deprecated)
Humanity's Last Exam	47.3%	Extra highGPT-5.6 Sol (Xhigh)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-xhigh) (AA marks this variant deprecated)
Humanity's Last Exam	49.5%	MaxGPT-5.6 Sol (Max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol) (AA marks this variant deprecated)
IOI (Vals)	91.2%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	29 Sep 2026	—	±4.51 stderr; $7.982/test
LegalBench (Vals)	87.0%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	29 Sep 2026	—	vals id openai/gpt-5.6-sol; rank 9/149; ±0.412 stderr; $0.012678/test
LLM Creative Story-Writing Benchmark (Lech Mazur)	2.3	HighGPT-5.6 Sol (high)	Independent testIndependentLech Mazur (LLM Creative Story-Writing Benchmark) ↗	28 Sep 2026	—	rank 11/56; Thurstone comparison score (centered at 0); est. win chance 81%; 95% bootstrap 2.217 to 2.368
LLM Creative Story-Writing Benchmark (Lech Mazur)	2.5	Extra highGPT-5.6 Sol (xhigh)	Independent testIndependentLech Mazur (LLM Creative Story-Writing Benchmark) ↗	28 Sep 2026	—	rank 8/56; Thurstone comparison score (centered at 0); est. win chance 83%; 95% bootstrap 2.407 to 2.561
LMArena Code Arena (WebDev)	1619	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	30 Sep 2026	Codex	Code Arena \| WebDev overall (agentic web-dev, raw); rank 22 (CI rank 15-24); 95% CI 1613-1625; 16911 votes
LMArena Search Arena	1257	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	24 Aug 2026	—	rank 1 (rank range 1-2); 95% CI 1249.8-1264.7; 29663 votes
LMArena Text - Coding category	1534	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	30 Sep 2026	—	text arena coding category, style control on; rank 14 (CI rank 6-33); 95% CI 1527-1541; 9788 votes
LMArena Text - Creative Writing	1469	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	30 Sep 2026	—	rank 19 (rank range 7-35); 95% CI 1460.9-1477.0; 7388 votes; style-controlled
LMArena Text - Expert	1535	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	30 Sep 2026	—	rank 6 (rank range 1-23); 95% CI 1524.6-1544.4; 4053 votes; style-controlled
LMArena Text - Hard Prompts	1511	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	30 Sep 2026	—	rank 16 (rank range 7-32); 95% CI 1505.6-1516.0; 23609 votes; style-controlled
LMArena Text - Instruction Following	1491	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	30 Sep 2026	—	rank 11 (rank range 6-22); 95% CI 1484.2-1497.0; 12619 votes; style-controlled
LMArena Text - Longer Query	1498	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	30 Sep 2026	—	rank 18 (rank range 7-31); 95% CI 1491.9-1503.9; 16784 votes; style-controlled
LMArena Text - Non-English	1475	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	30 Sep 2026	—	rank 20 (rank range 8-34); 95% CI 1469.2-1479.8; 21430 votes; style-controlled
LMArena Text - Occupational: Business, Management & Finance	1485	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	30 Sep 2026	—	rank 20 (rank range 6-43); 95% CI 1476.5-1492.5; 6778 votes; style-controlled
LMArena Text - Occupational: Legal & Government	1494	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	30 Sep 2026	—	rank 20 (rank range 2-57); 95% CI 1482.9-1505.7; 3029 votes; style-controlled
LMArena Text - Occupational: Writing, Literature & Language	1482	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	30 Sep 2026	—	rank 12 (rank range 6-28); 95% CI 1474.8-1489.1; 9675 votes; style-controlled
LMArena Text (overall)	1484	Extra highgpt-5.6-sol-xhigh	Independent testIndependentLMArena ↗	30 Sep 2026	—	text overall, style control on; rank 20 (CI rank 11-34); 95% CI 1479-1488; 36002 votes
MCP Atlas	81.8%	Defaultgpt-5.6 (sol)	Independent testIndependentScale AI SEAL ↗	1 Oct 2026	—	rank 2 (Scale rank accounts for CI); ±2.4; entry added 2026-07-15
MMMU-Pro (no tools)	83.0%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
MMMU-Pro (with tools)	84.6%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
OpenAI MRCR v2 (8-needle)	91.5%	Defaultbest score across efforts (OpenAI table: "maximum at any effort"); 8-needle 256K-512K	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—
OpenAI MRCR v2 (8-needle)	91.5%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results); 8-needle 256K-512K	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
OpenAI MRCR v2 (8-needle)	73.8%	Defaultbest score across efforts (OpenAI table: "maximum at any effort"); 8-needle 512K-1M	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—
OpenAI MRCR v2 (8-needle)	73.8%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results); 8-needle 512K-1M	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
OSWorld 2.0	27.3%	Maxgpt-5.6-sol_max	Independent testIndependentOSWorld 2.0 via Epoch AI ↗	1 Oct 2026	—	read from Epoch AI benchmark_data.zip (osworld_2_external.csv); original leaderboard https://osworld-v2.xlang.ai/; partial score 0.6272; tool setting batch tool; step budget 500
OSWorld 2.0	62.6%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—	OSWorld 2.0 (OpenAI-run)
OSWorld 2.0 offline set	16.4%	No reasoningreasoning effort=none	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	OSWorld 2.0 offline set, partial reward, v2026.08.08 release; vendor-estimated API cost/task $0.78171
OSWorld 2.0 offline set	29.8%	Lowreasoning effort=low	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	OSWorld 2.0 offline set, partial reward, v2026.08.08 release; vendor-estimated API cost/task $0.9086
OSWorld 2.0 offline set	49.7%	Mediumreasoning effort=medium	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	OSWorld 2.0 offline set, partial reward, v2026.08.08 release; vendor-estimated API cost/task $2.7275
OSWorld 2.0 offline set	56.5%	Highreasoning effort=high	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	OSWorld 2.0 offline set, partial reward, v2026.08.08 release; vendor-estimated API cost/task $4.4613
OSWorld 2.0 offline set	60.9%	Extra highreasoning effort=xhigh	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	OSWorld 2.0 offline set, partial reward, v2026.08.08 release; vendor-estimated API cost/task $5.9257
OSWorld 2.0 offline set	66.2%	Maxreasoning effort=max	Maker's own figureVendor-reportedOpenAI ↗	22 Sep 2026	—	OSWorld 2.0 offline set, partial reward, v2026.08.08 release; vendor-estimated API cost/task $7.7105
PRBench Finance (Scale)	50.5%	Maxgpt-5.6-sol (max)	Independent testIndependentScale AI (SEAL) ↗	28 Jul 2026	—	Scale rank 7; ±0.33 CI
PRBench Legal (Scale)	50.5%	Maxgpt-5.6-sol (max)	Independent testIndependentScale AI (SEAL) ↗	28 Jul 2026	—	Scale rank 7; ±0.54 CI
ProgramBench (avg test pass rate)	69.9%	Extra high	Independent testIndependentProgramBench ↗	28 Sep 2026	mini-SWE-agent	avg behavioral-test pass rate ("Score" column); fully resolved 1.0%, almost (>=95% tests) 15.5%; avg cost $6.08/task
ProgramBench (avg test pass rate)	57.8%	Default	Independent testIndependentProgramBench ↗	28 Sep 2026	mini-SWE-agent	avg behavioral-test pass rate ("Score" column); fully resolved 0.5%, almost (>=95% tests) 2.5%; avg cost $1.00/task
ProgramBench (fully resolved)	1.0%	Extra high	Independent testIndependentProgramBench ↗	28 Sep 2026	mini-SWE-agent	strict fully-resolved rate; almost (>=95% tests) 15.5%; avg cost $6.08/task
ProgramBench (fully resolved)	1.5%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	29 Sep 2026	mini-SWE-agent	±0.862 stderr; $15.289/test; strict fully-resolved rate
ProgramBench (fully resolved)	0.5%	Default	Independent testIndependentProgramBench ↗	28 Sep 2026	mini-SWE-agent	strict fully-resolved rate; almost (>=95% tests) 2.5%; avg cost $1.00/task
SciCode	56.4%	LowGPT-5.6 Sol (Low)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-low) (AA marks this variant deprecated)
SciCode	57.4%	MediumGPT-5.6 Sol (Medium)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-medium) (AA marks this variant deprecated)
SciCode	57.8%	HighGPT-5.6 Sol (High)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-high) (AA marks this variant deprecated)
SciCode	57.1%	Extra highGPT-5.6 Sol (Xhigh)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-xhigh) (AA marks this variant deprecated)
SciCode	57.1%	MaxGPT-5.6 Sol (Max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol) (AA marks this variant deprecated)
ScreenSpot-Pro	59.3%	Lowreasoning effort=low	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	ScreenSpot-Pro, no tools; vendor-estimated API cost/task $0.03213
ScreenSpot-Pro	68.7%	Mediumreasoning effort=medium	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	ScreenSpot-Pro, no tools; vendor-estimated API cost/task $0.03398
ScreenSpot-Pro	75.1%	Highreasoning effort=high	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	ScreenSpot-Pro, no tools; vendor-estimated API cost/task $0.03648
ScreenSpot-Pro	76.9%	Extra highreasoning effort=xhigh	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	ScreenSpot-Pro, no tools; vendor-estimated API cost/task $0.04098
ScreenSpot-Pro	76.8%	Maxreasoning effort=max	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	—	ScreenSpot-Pro, no tools; vendor-estimated API cost/task $0.04931
SimpleBench	64.8%	Extra highGPT-5.6 Sol (xhigh)	Independent testIndependentSimpleBench ↗	9 Jul 2026	—	AVG@5, temp 0.7; rank 24th; human baseline 83.7%; data https://simple-bench.com/static/js/leaderboard-data.js
SimpleQA Verified	69.7%	Maxgpt-5.6-sol_max	Independent testIndependentEpoch AI ↗	10 Aug 2026	—	Epoch-run (no tools); ±1.45 stderr
SkillsBench	54.1%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	27 Sep 2026	OpenHands	±4.73 stderr; $4.140/test
SWE Atlas - Codebase QnA	46.0%	Extra highGPT-5.6-Sol (Codex) xHigh*	Independent testIndependentScale AI SEAL ↗	1 Oct 2026	Codex	rank 2 (Scale rank accounts for CI); ±5; entry added 2026-03-02; * = Scale footnote (see page)
SWE Atlas - Codebase QnA	54.0%	MaxGPT-5.6 Sol (max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	Codex	AA Coding Agent Index component; pass@1 avg of 3 attempts
SWE Atlas - Test Writing	45.9%	Extra highGPT-5.6-Sol (Codex) xHigh*	Independent testIndependentScale AI SEAL ↗	1 Oct 2026	Codex	rank 1 (Scale rank accounts for CI); ±6; entry added 2026-07-28; * = Scale footnote (see page)
SWE-Bench Pro (public, v1)	64.6%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
SWE-Bench Pro V2 (full)	95.5%	Extra highGPT-5.6-sol (Codex) xhigh	Independent testIndependentScale AI SEAL ↗	1 Oct 2026	Codex	rank 6 (Scale rank accounts for CI); ±1.4; entry added 2026-09-22; SWE-Bench Pro V2 (642 tasks, locked protocol, released 2026-09-22)
SWE-Bench Pro V2 (hard)	82.4%	Extra highGPT-5.6 Sol (Codex) xhigh	Independent testIndependentScale AI SEAL ↗	1 Oct 2026	Codex	rank 8 (Scale rank accounts for CI); ±0; entry added 2026-09-22; SWE-Bench Pro V2 (642 tasks, locked protocol, released 2026-09-22)
SWE-bench Verified	96.2%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	1 Sep 2026	bash-only (single bash tool) agent	±0.856 stderr; $1.151/test; Vals archived SWE-bench Verified on 2026-09-01 (saturated)
SWE-rebench	62.3%	MediumGPT-5.6 Sol [medium]	Independent testIndependentSWE-rebench (Nebius) ↗	1 Oct 2026	SWE-rebench standard scaffold	time window 2026-05-15..2026-07-01 (111 problems, 65 repos); ±1.83; pass@5 79.3%; $0.85/problem
tau2-bench	76.0%	LowGPT-5.6 Sol (Low)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-low) (AA marks this variant deprecated)
tau2-bench	81.0%	MediumGPT-5.6 Sol (Medium)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-medium) (AA marks this variant deprecated)
tau2-bench	83.3%	HighGPT-5.6 Sol (High)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-high) (AA marks this variant deprecated)
tau2-bench	84.8%	Extra highGPT-5.6 Sol (Xhigh)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-xhigh) (AA marks this variant deprecated)
tau2-bench	85.1%	MaxGPT-5.6 Sol (Max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol) (AA marks this variant deprecated)
Terminal-Bench 2.1	76.8%	LowGPT-5.6 Sol (Low)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-low) (AA marks this variant deprecated)
Terminal-Bench 2.1	86.1%	MediumGPT-5.6 Sol (Medium)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-medium) (AA marks this variant deprecated)
Terminal-Bench 2.1	87.3%	HighGPT-5.6 Sol (High)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-high) (AA marks this variant deprecated)
Terminal-Bench 2.1	89.5%	Extra highGPT-5.6 Sol (Xhigh)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-xhigh) (AA marks this variant deprecated)
Terminal-Bench 2.1	88.0%	MaxGPT-5.6 Sol (Max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol) (AA marks this variant deprecated)
Terminal-Bench 2.1	85.8%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	27 Sep 2026	Terminus 2	±1.35 stderr; $1.023/test
Terminal-Bench 2.1	91.9%	DefaultGPT-5.6 Sol Ultra (4 parallel agents)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	Codex	ultra = multi-agent parallel test-time compute (4 agents by default)
Terminal-Bench 2.1	88.8%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	Codex	Terminal-Bench 2.1
Terminal-Bench 3.0	34.6%	Max	Independent testIndependentSnorkel AI / Terminal-Bench ↗	1 Oct 2026	Codex	TB 3.0 v0.1 (74 tasks); superseded by TB 4.0; tokens 5.8B, run cost $4.0k
Terminal-Bench 4.0	1.0%	LowGPT-5.6 Sol (Low)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-low) (AA marks this variant deprecated)
Terminal-Bench 4.0	7.9%	Lowreasoning effort=low	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	Codex	Terminal-Bench 4.0; vendor-estimated API cost/task $1.46089
Terminal-Bench 4.0	14.6%	MediumGPT-5.6 Sol (Medium)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-medium) (AA marks this variant deprecated)
Terminal-Bench 4.0	20.9%	Mediumreasoning effort=medium	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	Codex	Terminal-Bench 4.0; vendor-estimated API cost/task $2.69056
Terminal-Bench 4.0	20.7%	HighGPT-5.6 Sol (High)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-high) (AA marks this variant deprecated)
Terminal-Bench 4.0	26.1%	Highreasoning effort=high	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	Codex	Terminal-Bench 4.0; vendor-estimated API cost/task $4.12252
Terminal-Bench 4.0	24.7%	Extra highGPT-5.6 Sol (Xhigh)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol-xhigh) (AA marks this variant deprecated)
Terminal-Bench 4.0	28.5%	Extra highreasoning effort=xhigh	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	Codex	Terminal-Bench 4.0; vendor-estimated API cost/task $5.38518
Terminal-Bench 4.0	39.9%	MaxGPT-5.6 Sol (Max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA-run evaluation (AA slug gpt-5-6-sol) (AA marks this variant deprecated)
Terminal-Bench 4.0	37.9%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	29 Sep 2026	mini-SWE-agent	±0 stderr; $7.984/test
Terminal-Bench 4.0	37.4%	MaxGPT-5.6 Sol (max)	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	Codex	AA Coding Agent Index component run in the vendor agent harness; pass@1 avg of 3 attempts
Terminal-Bench 4.0	37.3%	Max	Independent testIndependentTerminal-Bench ↗	1 Oct 2026	Codex	TB 4.0.0 (66 tasks); ±3.78 95% CI; 330 trials; total run cost $2542; model release 2026-06-26
Terminal-Bench 4.0	37.3%	Maxreasoning effort=max	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	Codex	Terminal-Bench 4.0; vendor-estimated API cost/task $7.89451
Terminal-Bench-Science 0.1	22.4%	Maxreasoning effort=max	Maker's own figureVendor-reportedOpenAI ↗	3 Sep 2026	Codex	Terminal-Bench Science 0.1; vendor-estimated API cost/task $20.10413
Toolathlon	58.0%	DefaultGPT-5.6 launch table (effort not stated; text cites max for headline results)	Maker's own figureVendor-reportedOpenAI ↗	9 Jul 2026	—
Vals Code Migration	52.9%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	29 Sep 2026	—	±4.353 stderr; $24.541/test
Vals Index	58.0%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	30 Sep 2026	—	±1.017 stderr; $14.239/test
Vals Legal Research Bench	48.1%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	29 Sep 2026	—	vals id openai/gpt-5.6-sol; rank 9/72; ±3.473 stderr; $21.605566/test
Vals Public Benefits Bench v1.1	66.5%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	29 Sep 2026	—	vals id openai/gpt-5.6-sol; rank 16/45; ±1.228 stderr; $8.847882/test
Vals SRE Bench	30.5%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	29 Sep 2026	—	±2.851 stderr; $42.521/test
Vals Vibe Code Bench	80.5%	Maxreasoning_effort=max	Independent testIndependentVals.ai ↗	29 Sep 2026	OpenHands	±3.716 stderr; $33.404/test
Vectara Hallucination Leaderboard (HHEM)	12.4%	Defaultopenai/gpt-5.6-sol	Independent testIndependentVectara ↗	22 Sep 2026	—	factual consistency 87.6 %; answer rate 99.1 %; avg summary 131.1 words; HHEM-2.3 judge; effort not stated (API default)
Vending-Bench 2	$9,619	Default	Independent testIndependentAndon Labs ↗	1 Oct 2026	—	final money balance after simulated year, arithmetic mean across runs; ±$1,338; rank 7; only top 10 rendered server-side (57 more behind 'Show more')