Models · Cohere · Out sinceReleased 20 May 2026
Command A+
Command A+ is made by Cohere. We don't have enough test results yet to rank it. It's cheap to use. Its makers have published it, so you can run it on your own servers.
218B total / 25B active MoE, Apache 2.0 (Cohere's first fully Apache-licensed model), 48 languages incl. Swedish, Danish, Norwegian, Finnish, Icelandic. Card: 128K input context (config max_position_embeddings 200,000; OpenRouter lists 192K). BF16 / FP8 / W4A4 checkpoints; W4A4 runs on 2x H100 or 1x B200. Native citation grounding. OpenRouter ~$0.30/$1.50 (listed 2026-09-22); Cohere list price not verified.
- —
- —
- —
- Cheap$0.30 / $1.50
- $0.60
- 128K
- —
- 8 (8 independent8 indep.)
- 20 May 2026
- Cohere
- 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.
You can't choose how long this one thinks.
No adjustable reasoning setting listed.
OpenRouter is a service that gives access to many AI models in one place. This model is listed there as cohere/command-a-plus.
Route it as cohere/command-a-plus at $0.30 in / $1.50 out per 1M tokens, 192K context. Listed since 22 Sep 2026.
frequency_penaltyinclude_reasoningmax_tokenspresence_penaltyreasoningresponse_formatseedstopstructured_outputstemperaturetoolstop_ktop_p
Run it yourself
Self-hosting
On your own servers,Licence, sizeunder your own control.and hardware.
- Yes
- 218B / 25B
- No
Needs one full AI server.
≤ 1.1 TB at 8-bit (one 8×H200 node). Weights ≈ 229 GB at 8-bit, 120 GB at 4-bit (+10–30% for KV cache). MoE: 25B active per token.
Languages: Languages: Trained on 48 languages incl. Swedish, Danish, Norwegian, Finnish, Icelandic (card).
Training it further: Fine-tuning: No recipe in card.
Quantisations: bf16, fp8 (CohereLabs/command-a-plus-05-2026-fp8), w4a4 (CohereLabs/command-a-plus-05-2026-w4a4, vendor-recommended), gguf (community: bartowski). Engines: vLLM (>=0.21 + cohere_melody parser), Transformers.
Download (Hugging Face)Weights on Hugging Face ↗huggingface.co ↗cohere.com ↗
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.
| NotesNotes | ||||||
|---|---|---|---|---|---|---|
| AA-LCR | 52.7% | DefaultCommand A+ | Independent testIndependentArtificial Analysis ↗ | 1 Oct 2026 | — | AA slug command-a-plus |
| AA-Omniscience Index | -4 | DefaultCommand A+ | Independent testIndependentArtificial Analysis ↗ | 1 Oct 2026 | — | AA slug command-a-plus; AA-Omniscience Index (-100..100) |
| Artificial Analysis Intelligence Index | 13.1 | DefaultCommand A+ | Independent testIndependentArtificial Analysis ↗ | 1 Oct 2026 | — | AA slug command-a-plus |
| Artificial Analysis output speed | 208 tok/s | DefaultCommand A+ | Independent testIndependentArtificial Analysis ↗ | 1 Oct 2026 | — | AA slug command-a-plus; median output tokens/sec across AA-tracked API providers (not a self-hosted measurement) |
| GPQA Diamond | 76.1% | DefaultCommand A+ | Independent testIndependentArtificial Analysis ↗ | 1 Oct 2026 | — | AA slug command-a-plus |
| Humanity's Last Exam | 12.0% | DefaultCommand A+ | Independent testIndependentArtificial Analysis ↗ | 1 Oct 2026 | — | AA slug command-a-plus |
| SciCode | 38.5% | DefaultCommand A+ | Independent testIndependentArtificial Analysis ↗ | 1 Oct 2026 | — | AA slug command-a-plus |
| Terminal-Bench 2.1 | 22.9% | DefaultCommand A+ | Independent testIndependentArtificial Analysis ↗ | 1 Oct 2026 | — | AA slug command-a-plus |