whatmodel.sh / Command A+ (05-2026)
Command A+ (05-2026)
↗ share
+ contribute
Command A+ (05-2026) is an open-weights language model from Cohere, released May 2026. Its published size is 218B parameters (25B active per token) and a 128k-token context window. Licence: apache-2.0.
$ how it compares
The LLMs ranked either side of Command A+ (05-2026), out of 828 with a score, then more from Cohere.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 173 |  Qwen3 235B-A22B Instruct 2507 | 65±1 | api $0.20/M +local | ||
| 174 |  Mistral Medium 3.5 (128B) | 65±1 | api $3.00/M +local | ||
| 175 | Command A+ (05-2026) | 64±1 | api price unknown +local | ||
| 176 |  Claude Haiku 4.5 | 64±1 | api $2.00/M | ||
| 177 |  Nemotron-Labs-3 Puzzle 75B-A9B | 64±1 | local | ||
| more from Cohere | |||||
| 76 | North Small Translate 1.0 | 71±6 | api price unknown | ||
| 277 | Command A (03-2025) | 57±1 | api $4.38/M +local | ||
| 324 | Command A Reasoning (08-2025) | 55±2 | local | ||
$ what it's good at
agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #138 of 300 · 25% of its score · 6 tests
Beats DeepSeek V3.2 Exp, Qwen3 235B-A22B Thinking 2507, o4-mini here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Large, GPT-4o (2024-05-13) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Terminal-Bench Hard +1 | 25.0% ✎ | ▄········██▄ beats 1/5 | 8.6% |
| τ²-bench Telecom lab | 85.0% ✎ | ▁·▅█▆▅▂▃▃▂▃▁ beats 63/89 | 5.9% |
| APEX-Agents (pass@1) | 1.6% ✎ | ▄▄····▄···█▄ beats 0/5 | 3.9% |
| MCP-Atlas (public set) | 27.2% ✎ | ▄▄▄·······█▄ beats 0/5 | 3.4% |
| SWE-bench Verified | 14.4% ✎ | ▄··········█ beats 0/5 | 2.5% |
| τ³-bench Banking | 5.8% ✎ | █·▄▄······▄▄ beats 1/5 | 1.0% |
coding▁▁▁▃▆██▇▆▄▃▁ #152 of 458 · 9.2% of its score · 3 tests
Beats Gemini 2.5 Pro, Gemini 3.1 Flash-Lite, Qwen3-Coder-Next (80B-A3B) here, though they rank higher overall. Trails Ministral 3 8B Reasoning 2512, Phi-4-reasoning 14B, DeepSeek V3 (Dec '24) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LiveCodeBench v6 (window unspecified) | 86.1% ✎ | ▄······▄▄▄·█ beats 2/5 | 4.3% |
| SWE-bench Verified | 14.4% ✎ | ▄··········█ beats 0/5 | 2.5% |
| SciCode (level unspecified) lab | 38.0% ✎ | ▁····▃▄▄▄█▆▂ beats 21/34 | 2.4% |
math▁▂▃▅▆▅▄█▄▁▁▁ #104 of 398 · 17% of its score · 3 tests
Beats Claude Sonnet 4.6, MiniMax-M3, Nemotron 3 Ultra 550B-A55B here, though they rank higher overall. Trails Gemini 2.5 Flash, gpt-oss-120b, GLM-4.5-Air here, though they rank lower overall.
vision & documents······█····· #1 of 1 * · 5.8% of its score · 2 tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| CharXiv (reasoning questions) lab | 52.7% ✎ | ▂▅▃▅▅▆▆▅▆█▆▂ beats 32/100 | 2.9% |
| CharXiv descriptive questions lab | 88.0% ✎ | ▂▄▄▅▇▄█▇▇▂▁▂ beats 42/58 | 2.9% |
other tests9 not in a task · 45% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GPQA Diamond | 75.6% ✎ | █···██·█··██ beats 1/5 | 20% |
| MMLU-Pro | 79.0% ✎ | ▄·▄▄····▄··█ beats 0/5 | 6.5% |
| Humanity's Last Exam | 11.4% ✎ | █▄······▄·▄▄ beats 0/5 | 6.1% |
| IFBench (aggregation unspecified) | 73.9% ✎ | ▄···▄▄··▄··█ beats 3/5 | 4.2% |
| MultiChallenge | 45.8% ✎ | █··▄·▄·····█ beats 2/5 | 3.3% |
| WMT25 Terminology Translation (term success rate) lab | 68.7% ✎ | ▂·▃··▂·▄▂·█▆ beats 1/12 | 1.7% |
| IFBench prompt-level (loose) lab | 74.0% ✎ | ▃········██▅ beats 2/8 | 1.5% |
| WMT24++ lab | 81.0% ✎ | █·▃·▅▅▃···▃▃ beats 10/10 | 0.8% |
1 more test
| test | result | vs the others in its run | counts |
|---|---|---|---|
| WMT26 General MT (GEMBA-ESA, LLM judge) lab | 76.5% ✎ | ▄·▄▄▄·▄██▄█▄ beats 6/12 | 0.6% |
$ where to run it
Through an API
api · price unknown · 128k context · docs.cohere.com ↗
llama.cpp · recommended
version
- Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.macOS / Linux
curl -LsSf https://llama.app/install.sh | shWindowsirm https://llama.app/install.ps1 | iex - Run Command A+ (05-2026) · needs ~131.3 GB · from huggingface.co/bartowskichat in the terminal
llama cli -hf bartowski/command-a-plus-05-2026-GGUF:Q4_K_M - Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf bartowski/command-a-plus-05-2026-GGUF:Q4_K_M
LM Studio
version
- Install LM Studio · Open the app once before using the lms command.
- Run Command A+ (05-2026) · needs ~131.3 GB · from huggingface.co/bartowski
- Or download it in the terminal
lms get https://huggingface.co/bartowski/command-a-plus-05-2026-GGUF@Q4_K_M -y - Then run lms ls to see its name, and lms chat followed by that name
Weights and documentation: huggingface.co/CohereLabs ↗
$ how the score was worked out · Solar Open 2 Technical Report… 72%, Cohere vendor-reported benchm… 25%, Cohere — North Small Translat… 3%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 Solar Open 2 Technical Report (Upstage, arXiv:2607.20062v2) | 13 | 72% |
 Cohere vendor-reported benchmark scores for its own models (cohere.com blog) · lab-reported | 7 | 25% |
| Cohere — North Small Translate (arXiv:2609.13916v1) · lab-reported | 3 | 3% |
| prior · pull toward the middle | 0% | |
| weighted average | = 64.3 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json