CohereLanguage model
Command A Reasoning
Command A Reasoning has no overall index value yet because there are too few independent benchmark results.
PopularityRank 692 of 785
21out of 100-2 vs. last month
based on search, usage and attention
Summary
Command A Reasoning is Cohere's first Command checkpoint explicitly released as a reasoning model. It can think before answering or respond directly. Its 111 billion parameters, 256,000-token context, and 23 languages belong to a separate model identity alongside Command A.
Cohere lists it as live and releases the weights openly. The 32,000-token output cap includes visible reasoning content. Tool workflows therefore have to fit that reasoning trace and their tool-call history into the available context.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 73 published benchmark results
Page 1 of 7
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
GPQA Diamond (Show measurement, test conditions, and source)
- Source value
- 66.67
- Score
- 66.67
- Metric
- accuracy as shown in the evaluation result
- Unit
- %
- Benchmark version
- GPQA Diamond; exact evaluation date/settings not shown
- Category
- reasoning
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- Cohere Command A Reasoning model-card evaluation result
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Official benchmark leaderboard. Exact command-a-reasoning-08-2025 Hugging Face evaluation result, GPQA Diamond entry; the card does not state prompting, sampling or task date.
- Context
- The model card links the published evaluation result; no independent rerun or full protocol was established.
Terminal-Bench Hard (Show measurement, test conditions, and source)
- Source value
- 3
- Score
- 3
- Metric
- success score as reported
- Unit
- %
- Benchmark version
- Terminal-Bench Hard; Terminus-2; Artificial Analysis methodology
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Command A family comparison
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact Command A Reasoning comparison row; Terminal-Bench Hard uses Terminus-2 following Artificial Analysis methodology.
- Context
- Cohere reports scores; exact task snapshot and all inference settings are not disclosed.
τ²-Bench Telecom (Show measurement, test conditions, and source)
- Source value
- 37
- Score
- 37
- Metric
- task completion score as reported
- Unit
- %
- Benchmark version
- τ²-Bench; standard settings and user simulation, Barres et al. 2025
- Category
- agentic
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere Command A family comparison
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact Command A Reasoning comparison row; τ²-Bench Telecom standard settings and user simulation, Barres et al. 2025.
- Context
- Cohere reports the value; exact run count and all configuration details are not disclosed.
K-MetBench overall accuracy (Show measurement, test conditions, and source)
- Source value
- 77.8
- Score
- 77.8
- Metric
- accuracy
- Unit
- %
- Benchmark version
- ACL 2026 Findings Table 4, full benchmark
- Category
- reasoning
- Direction
- Higher is better
- Sample
- 1774
- Source type
- independent-evaluator
- Evaluator
- K-MetBench: A Multi-Dimensional Benchmark for Fine-Grained Evaluation of Expert Reasoning, Locality, and Multimodality in Meteorology (ACL 2026 Findings)
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact command-a-reasoning-08-2025 row; zero-shot standard prompt in original Korean; vLLM defaults, temperature 1.0 for reasoning models, seed 42; answer-format violations count as failures.
- Context
- Narrow Korean meteorology evaluation; not a general capability score. The paper names this exact model ID but does not provide an immutable API or weight revision.
K-MetBench Korean geo-cultural accuracy (Show measurement, test conditions, and source)
- Source value
- 74.6
- Score
- 74.6
- Metric
- accuracy
- Unit
- %
- Benchmark version
- ACL 2026 Findings Table 4, Korean-specific subset
- Category
- reasoning
- Direction
- Higher is better
- Sample
- 73
- Source type
- independent-evaluator
- Evaluator
- K-MetBench: A Multi-Dimensional Benchmark for Fine-Grained Evaluation of Expert Reasoning, Locality, and Multimodality in Meteorology (ACL 2026 Findings)
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact model row; Korean-specific geo-cultural subset; zero-shot standard prompt in original Korean, reasoning-model temperature 1.0, seed 42.
- Context
- Narrow Korean meteorology evaluation; not a general capability score. The paper names this exact model ID but does not provide an immutable API or weight revision.
K-MetBench text-only modality accuracy (Show measurement, test conditions, and source)
- Source value
- 77.8
- Score
- 77.8
- Metric
- accuracy
- Unit
- %
- Benchmark version
- ACL 2026 Findings Table 4, text-only modality subset
- Category
- reasoning
- Direction
- Higher is better
- Sample
- 1692
- Source type
- independent-evaluator
- Evaluator
- K-MetBench: A Multi-Dimensional Benchmark for Fine-Grained Evaluation of Expert Reasoning, Locality, and Multimodality in Meteorology (ACL 2026 Findings)
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact model row; text-only questions, excluding the 82 image-plus-text questions; zero-shot standard prompt in Korean, reasoning-model temperature 1.0, seed 42.
- Context
- Narrow Korean meteorology evaluation; not a general capability score. The paper names this exact model ID but does not provide an immutable API or weight revision.
K-MetBench Part 1 Forecast Theory accuracy (Show measurement, test conditions, and source)
- Source value
- 73.4
- Score
- 73.4
- Metric
- accuracy
- Unit
- %
- Benchmark version
- ACL 2026 Findings Table 4, Korean meteorology certification subject area
- Category
- reasoning
- Direction
- Higher is better
- Sample
- 373
- Source type
- independent-evaluator
- Evaluator
- K-MetBench: A Multi-Dimensional Benchmark for Fine-Grained Evaluation of Expert Reasoning, Locality, and Multimodality in Meteorology (ACL 2026 Findings)
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact model row; Forecast Theory subset of 373 questions; zero-shot standard prompt in Korean, reasoning-model temperature 1.0, seed 42.
- Context
- Narrow Korean meteorology evaluation; not a general capability score. The paper names this exact model ID but does not provide an immutable API or weight revision.
K-MetBench Part 2 Meteorological Observation Methods accuracy (Show measurement, test conditions, and source)
- Source value
- 85.2
- Score
- 85.2
- Metric
- accuracy
- Unit
- %
- Benchmark version
- ACL 2026 Findings Table 4, Korean meteorology certification subject area
- Category
- reasoning
- Direction
- Higher is better
- Sample
- 332
- Source type
- independent-evaluator
- Evaluator
- K-MetBench: A Multi-Dimensional Benchmark for Fine-Grained Evaluation of Expert Reasoning, Locality, and Multimodality in Meteorology (ACL 2026 Findings)
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact model row; Meteorological Observation Methods subset of 332 questions; zero-shot standard prompt in Korean, reasoning-model temperature 1.0, seed 42.
- Context
- Narrow Korean meteorology evaluation; not a general capability score. The paper names this exact model ID but does not provide an immutable API or weight revision.
K-MetBench Part 3 Atmospheric Dynamics accuracy (Show measurement, test conditions, and source)
- Source value
- 73.8
- Score
- 73.8
- Metric
- accuracy
- Unit
- %
- Benchmark version
- ACL 2026 Findings Table 4, Korean meteorology certification subject area
- Category
- reasoning
- Direction
- Higher is better
- Sample
- 359
- Source type
- independent-evaluator
- Evaluator
- K-MetBench: A Multi-Dimensional Benchmark for Fine-Grained Evaluation of Expert Reasoning, Locality, and Multimodality in Meteorology (ACL 2026 Findings)
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact model row; Atmospheric Dynamics subset of 359 questions; zero-shot standard prompt in Korean, reasoning-model temperature 1.0, seed 42.
- Context
- Narrow Korean meteorology evaluation; not a general capability score. The paper names this exact model ID but does not provide an immutable API or weight revision.
K-MetBench Part 4 Climatology accuracy (Show measurement, test conditions, and source)
- Source value
- 78.8
- Score
- 78.8
- Metric
- accuracy
- Unit
- %
- Benchmark version
- ACL 2026 Findings Table 4, Korean meteorology certification subject area
- Category
- reasoning
- Direction
- Higher is better
- Sample
- 376
- Source type
- independent-evaluator
- Evaluator
- K-MetBench: A Multi-Dimensional Benchmark for Fine-Grained Evaluation of Expert Reasoning, Locality, and Multimodality in Meteorology (ACL 2026 Findings)
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact model row; Climatology subset of 376 questions; zero-shot standard prompt in Korean, reasoning-model temperature 1.0, seed 42.
- Context
- Narrow Korean meteorology evaluation; not a general capability score. The paper names this exact model ID but does not provide an immutable API or weight revision.
K-MetBench Part 5 Atmospheric Physics accuracy (Show measurement, test conditions, and source)
- Source value
- 78.5
- Score
- 78.5
- Metric
- accuracy
- Unit
- %
- Benchmark version
- ACL 2026 Findings Table 4, Korean meteorology certification subject area
- Category
- reasoning
- Direction
- Higher is better
- Sample
- 334
- Source type
- independent-evaluator
- Evaluator
- K-MetBench: A Multi-Dimensional Benchmark for Fine-Grained Evaluation of Expert Reasoning, Locality, and Multimodality in Meteorology (ACL 2026 Findings)
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact model row; Atmospheric Physics subset of 334 questions; zero-shot standard prompt in Korean, reasoning-model temperature 1.0, seed 42.
- Context
- Narrow Korean meteorology evaluation; not a general capability score. The paper names this exact model ID but does not provide an immutable API or weight revision.
Cohere North internal evaluation / Memory Usage Quality (Show measurement, test conditions, and source)
- Source value
- 39
- Score
- 39
- Metric
- Memory Usage Quality (LLM-as-a-judge score)
- Unit
- %
- Benchmark version
- Cohere North internal evaluation, reported in May 2026
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere, Introducing Command A+
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Exact Cohere model ID: command-a-reasoning-08-2025, evaluated within Cohere North on using information stored in a prior session to answer a question in a later session. Cohere labels this metric Memory Usage Quality and says it uses LLM-as-a-judge. The source does not disclose the prompt set, sampling protocol, generation parameters, evaluation N, or CI.
- Context
- Vendor-reported score from a proprietary Cohere North internal evaluation. The source supplies no public task data, sample count, confidence interval, or independently reproducible protocol. This is a North system-level observation, not an isolated model-only test.
Page 1 of 7
Popularity compared
How often the model is searched for, used and mentioned, compared with every model of its category.
Popularity index
The 100 most popular models of this category, 0 to 100, higher is more popular
20 of 101 models
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 7, 2026
Command A ReasoningGrok 4.3 22InternLM2-Chat-20B 22Pixtral Large (latest) 22Tiny Aya Water 21
Method changes in the timeline, a jump at these points is no market event:
- 1November 1, 2026: Value only with search or usage data, 30-day attention
What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.
Similar models
More models of the same kind as Command A Reasoning, chosen by model line, developer, size, price and release date.
- Command A VisionCohere · same developer
- Command A TranslateCohere · same developer
- Command ACohere · same developer
- Command A PlusCohere · same developer
- North-Small-Translate-1.0Cohere · same developer
- Aya Vision 32BCohere · same developer
- Aya Expanse 32BCohere · same developer
- North Mini CodeCohere · same developer
All specifications
Architecture
Input and output
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
Frequently asked questions about Command A Reasoning
What is the knowledge cutoff of Command A Reasoning?
June 2024, according to Cohere. That is 14 months before release.
Stated by Cohere
JunSepDecMarJunKnowledge cutoffJune 2024ReleasedAug 21, 2025
How large is the context window of Command A Reasoning?
256,000 tokens. That is how much text Command A Reasoning reads at once, about 380 A4 pages.
Page count as a rule of thumb from Artificial Analysis
Kimi K2.6262K
Qwen 3.8 27B262K
Kimi K2.7 Code262K
Kimi K2.5262K
Nemotron 3 Ultra262K
Kimi K2 Thinking262K
Command A ReasoningRank 77 of 103256K
Can I run Command A Reasoning locally?
Yes, the weights are open. With 111B parameters it needs a lot of memory, more a server than a laptop.
Stated by Cohere
- Open weights
- yes
- License
- CC BY-NC-4.0
- Parameters
- 111B
- active per token
- n/a
Sources and data date
Show all 19 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
| Evaluator | Test series | Composite indices | Recorded active measurements |
|---|---|---|---|
| Berkeley Function Calling Leaderboard | 0 | 0 | 1 |
Without vendor claims and historical or disputed results; those stay labelled in the single values. 1 measurements are not yet assigned to a test series.