Language modelDeprecated

Command R

Cohere

Released: March 2024
Data date: September 8, 2026

Input: Text
Output: Text

Command R is Cohere's original RAG and agent checkpoint from the first Command R release. The text model was built for multilingual enterprise work, retrieval-augmented generation, and tool use. It handles ten languages and up to 128,000 tokens. The open-weight release remains useful when you need a clearly documented RAG focus.

Cohere identifies this profile as command-r-03-2024 and marks it deprecated as of September 15, 2025. The command-r-08-2024 update is a separate model identity. Do not merge the two just because their API names look similar. CC-BY-NC-4.0 and the 4,000-token maximum output also constrain commercial and text-heavy use.

Profile

Specifications and access

Published information about this model. Existing estimates are explicitly labeled.

Dated documentation and model-card observations. Provider limits, native context and extended context can differ. Configuration notes retain the source wording.

Specification	Value and source
Weights	Open weightsSource not specified
Access	LocalSource
Parameters	32BSource not specified
Context window	128,000 tokensSource
License	CC-BY-NC-4.0Source
Release	2024-03-11Source
Parameters	35 BSource
Architecture	auto-regressive transformerSource
Maximum output	4,000 tokensSource
Input	TextSource
Output	TextSource
Languages	10Source
Tool calling	YesSource
API model ID	command-r-03-2024Source
Model revision release	2024-08-30 (This dates the separate command-r-08-2024 refresh and is not an original Command R release date.)Source

Published benchmarks

Legal

LegalBench

Sub-task	Command R (This model)	Command R+
Conclusion Tasks	13.98%	48.31% (best value in this row)
Interpretation Tasks	38.72%	71.62% (best value in this row)
Issue Tasks	8.89%	76.32% (best value in this row)
Rhetoric Tasks	71.01%	75.64% (best value in this row)
Rule Tasks	32.24%	69.55% (best value in this row)

higher is betterAs of August 11, 2026Source: Vals AIbars scaled within each row, not across rows

Methodology and context

Metric: accuracy

LegalBench

Overall

1.Claude Fable 588.56%

2.Claude Fable 5.188.51%

3.Gemini 3.1 Pro Preview87.4%

4.Gemini 3.7 Flash87.26%

5.Gemini 3.8 Flash86.99%

6.Claude Opus 586.97%

7.GPT-5.6 Sol86.97%

8.Gemini 3 Flash Preview86.86%

9.Gemini 3.6 Flash86.71%

10.GPT-5.586.52%

11.Grok 4.686.31%

12.GPT-5.486.04%

13.GPT-586.02%

14.Kimi K386.02%

15.Grok 4.585.97%

16.MiniMax M385.42%

17.Claude Opus 4.785.25%

18.GPT-5.6 Terra85.11%

19.Qwen 3.7 Max84.91%

20.GLM-5.384.84%

21.Kimi K2.684.74%

22.Grok 4.384.46%

23.GLM-5.184.39%

24.Grok 4.784.39%

25.Gemini 3.5 Flash-Lite84.07%

25.GLM-5.284.07%

27.GLM-584.06%

28.GPT-5.6 Luna84.03%

29.MiniMax M2.783.98%

30.GLM-5.3-Flash83.93%

31.Claude Sonnet 583.92%

32.Gemini 3.1 Flash-Lite83.76%

33.o383.76%

34.Qwen 3.8 Max 090283.61%

35.Gemini 3.5 Flash83.6%

36.Claude Opus 4.883.57%

37.GLM-4.783.36%

38.DeepSeek-V4.1-Flash83.28%

39.GPT-4.183.1%

40.Inkling Small82.99%

41.Inkling82.95%

42.Claude Opus 4.582.84%

43.Qwen 3.8 27B82.43%

44.Claude Sonnet 4.682.12%

45.Nemotron 3 Ultra82.07%

46.GPT-5 mini81.77%

47.Kimi K2 071181.45%

48.DeepSeek-V380.76%

49.o180.39%

50.DeepSeek-V4-Pro80.32%

51.Kimi K2 Thinking80.2%

52.MiniMax M2.579.96%

53.Command A79.7%

54.Ling 3.0 Flash79.69%

55.GLM-4.679.61%

56.o4-mini79.19%

57.Mistral Large 379.14%

58.MiMo-V2.578.89%

59.GPT-4.1 mini78.04%

60.GPT-5.4 nano77.92%

61.Grok 4.20 Reasoning77.74%

62.DeepSeek-V3 032477.73%

63.MiMo-V2.5-Pro77.13%

64.GPT OSS 120B75.94%

65.GLM-4.575.63%

66.MiniMax M2.175.45%

67.Llama 4 Scout72.04%

68.o3-mini71.54%

69.GPT OSS 20B70.85%

70.Mistral Small 3.169.16%

71.Command R+68.29%

72.Gemma 2 27B67.98%

73.DeepSeek-R167.32%

74.DeepSeek-V3.265.96%

75.GPT-4.1 nano61.06%

76.GPT-5 nano50.13%

77.Command R32.97%

higher is betterAs of September 29, 2026Scale: 0% - 100%

Source: Vals AI

CC BY 4.0

gradually.ai

Methodology and context

Comparison cohort: vals-legal-bench:1:overall
Metric: accuracy
Scale: 0-100
Participants: 77
Methodology and context: Source model ID: cohere/command-r. Source model name: Command R. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-09-27. Latency in seconds: 0.244. Source ID: vals-ai. Evidence capture ID: vals-legal_bench-2026-09-29-c1213019b552. Source payload hash: c1213019b55234a262ae86fd4ebfe291530f315c988a4ee32fd501c13777f85f. Parser version: vals-astro-v2. Imported at: 2026-09-29

Source: Vals AI

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.

Average (Show measurement, test conditions, and source)

Source value: 68.5
Score: 68.5
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:command-r:0
Source type: official-leaderboard
Evaluator: Open LLM leaderboard table in Cohere Command R+ card
Status: active
Retrieved at: 2026-09-08
Methodology: Official benchmark leaderboard. Original Command R row in standardized leaderboard table

Source Open LLM leaderboard table in Cohere Command R+ card

Arc (Challenge) (Show measurement, test conditions, and source)

Source value: 65.5
Score: 65.5
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:command-r:1
Source type: official-leaderboard
Evaluator: Open LLM leaderboard table in Cohere Command R+ card
Status: active
Retrieved at: 2026-09-08
Methodology: Official benchmark leaderboard. Original Command R row in standardized leaderboard table

Source Open LLM leaderboard table in Cohere Command R+ card

Hella Swag (Show measurement, test conditions, and source)

Source value: 87
Score: 87
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:command-r:2
Source type: official-leaderboard
Evaluator: Open LLM leaderboard table in Cohere Command R+ card
Status: active
Retrieved at: 2026-09-08
Methodology: Official benchmark leaderboard. Original Command R row in standardized leaderboard table

Source Open LLM leaderboard table in Cohere Command R+ card

MMLU (Show measurement, test conditions, and source)

Source value: 68.2
Score: 68.2
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:command-r:3
Source type: official-leaderboard
Evaluator: Open LLM leaderboard table in Cohere Command R+ card
Status: active
Retrieved at: 2026-09-08
Methodology: Official benchmark leaderboard. Original Command R row in standardized leaderboard table

Source Open LLM leaderboard table in Cohere Command R+ card

Truthful QA (Show measurement, test conditions, and source)

Source value: 52.3
Score: 52.3
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:command-r:4
Source type: official-leaderboard
Evaluator: Open LLM leaderboard table in Cohere Command R+ card
Status: active
Retrieved at: 2026-09-08
Methodology: Official benchmark leaderboard. Original Command R row in standardized leaderboard table

Source Open LLM leaderboard table in Cohere Command R+ card

Winogrande (Show measurement, test conditions, and source)

Source value: 81.5
Score: 81.5
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:command-r:5
Source type: official-leaderboard
Evaluator: Open LLM leaderboard table in Cohere Command R+ card
Status: active
Retrieved at: 2026-09-08
Methodology: Official benchmark leaderboard. Original Command R row in standardized leaderboard table

Source Open LLM leaderboard table in Cohere Command R+ card

Recorded measurement details

Every observation retains its source value and published test conditions.

LegalBench: Conclusion Tasks (Show measurement, test conditions, and source)

Source value: 13.98
Score: 13.98
Metric: accuracy
Unit: %
Benchmark version: 1
Category: legal
Task: conclusion_tasks
Task label: Conclusion Tasks
Direction: Higher is better
Comparison cohort: vals-legal-bench:1:conclusion_tasks
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-08-11
Methodology: Source model ID: cohere/command-r. Source model name: Command R. Provider reported by source: Cohere. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-08-09. Temperature: 0.3. Maximum output tokens: 4000. Latency in seconds: 0.168. Standard error: 0.699. Source ID: vals-ai. Evidence capture ID: vals-legal_bench-2026-08-11-836e5e44618e. Source payload hash: 836e5e44618e5bb1564702e615e2a5d9f70e26300758fa7cf8bad50033a889bb. Parser version: vals-astro-v2. Imported at: 2026-08-11
Comparison methodology: Temperature: 0.3. Maximum output tokens: 4000

Source Vals AI

LegalBench: Interpretation Tasks (Show measurement, test conditions, and source)

Source value: 38.716
Score: 38.716
Metric: accuracy
Unit: %
Benchmark version: 1
Category: legal
Task: interpretation_tasks
Task label: Interpretation Tasks
Direction: Higher is better
Comparison cohort: vals-legal-bench:1:interpretation_tasks
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-08-11
Methodology: Source model ID: cohere/command-r. Source model name: Command R. Provider reported by source: Cohere. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-08-09. Temperature: 0.3. Maximum output tokens: 4000. Latency in seconds: 0.186. Standard error: 0.208. Source ID: vals-ai. Evidence capture ID: vals-legal_bench-2026-08-11-836e5e44618e. Source payload hash: 836e5e44618e5bb1564702e615e2a5d9f70e26300758fa7cf8bad50033a889bb. Parser version: vals-astro-v2. Imported at: 2026-08-11
Comparison methodology: Temperature: 0.3. Maximum output tokens: 4000

Source Vals AI

LegalBench: Issue Tasks (Show measurement, test conditions, and source)

Source value: 8.888
Score: 8.888
Metric: accuracy
Unit: %
Benchmark version: 1
Category: legal
Task: issue_tasks
Task label: Issue Tasks
Direction: Higher is better
Comparison cohort: vals-legal-bench:1:issue_tasks
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-08-11
Methodology: Source model ID: cohere/command-r. Source model name: Command R. Provider reported by source: Cohere. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-08-09. Temperature: 0.3. Maximum output tokens: 4000. Latency in seconds: 0.22. Standard error: 0.357. Source ID: vals-ai. Evidence capture ID: vals-legal_bench-2026-08-11-836e5e44618e. Source payload hash: 836e5e44618e5bb1564702e615e2a5d9f70e26300758fa7cf8bad50033a889bb. Parser version: vals-astro-v2. Imported at: 2026-08-11
Comparison methodology: Temperature: 0.3. Maximum output tokens: 4000

Source Vals AI

LegalBench: Rhetoric Tasks (Show measurement, test conditions, and source)

Source value: 71.005
Score: 71.005
Metric: accuracy
Unit: %
Benchmark version: 1
Category: legal
Task: rhetoric_tasks
Task label: Rhetoric Tasks
Direction: Higher is better
Comparison cohort: vals-legal-bench:1:rhetoric_tasks
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-08-11
Methodology: Source model ID: cohere/command-r. Source model name: Command R. Provider reported by source: Cohere. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-08-09. Temperature: 0.3. Maximum output tokens: 4000. Latency in seconds: 0.163. Standard error: 1.375. Source ID: vals-ai. Evidence capture ID: vals-legal_bench-2026-08-11-836e5e44618e. Source payload hash: 836e5e44618e5bb1564702e615e2a5d9f70e26300758fa7cf8bad50033a889bb. Parser version: vals-astro-v2. Imported at: 2026-08-11
Comparison methodology: Temperature: 0.3. Maximum output tokens: 4000

Source Vals AI

LegalBench: Rule Tasks (Show measurement, test conditions, and source)

Source value: 32.235
Score: 32.235
Metric: accuracy
Unit: %
Benchmark version: 1
Category: legal
Task: rule_tasks
Task label: Rule Tasks
Direction: Higher is better
Comparison cohort: vals-legal-bench:1:rule_tasks
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-08-11
Methodology: Source model ID: cohere/command-r. Source model name: Command R. Provider reported by source: Cohere. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-08-09. Temperature: 0.3. Maximum output tokens: 4000. Latency in seconds: 0.485. Standard error: 3.169. Source ID: vals-ai. Evidence capture ID: vals-legal_bench-2026-08-11-836e5e44618e. Source payload hash: 836e5e44618e5bb1564702e615e2a5d9f70e26300758fa7cf8bad50033a889bb. Parser version: vals-astro-v2. Imported at: 2026-08-11
Comparison methodology: Temperature: 0.3. Maximum output tokens: 4000

Source Vals AI

LegalBench: Overall (Show measurement, test conditions, and source)

Source value: 32.965
Score: 32.965
Metric: accuracy
Unit: %
Benchmark version: 1
Category: legal
Task: overall
Task label: Overall
Direction: Higher is better
Comparison cohort: vals-legal-bench:1:overall
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-09-29
Methodology: Source model ID: cohere/command-r. Source model name: Command R. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-09-27. Latency in seconds: 0.244. Source ID: vals-ai. Evidence capture ID: vals-legal_bench-2026-09-29-c1213019b552. Source payload hash: c1213019b55234a262ae86fd4ebfe291530f315c988a4ee32fd501c13777f85f. Parser version: vals-astro-v2. Imported at: 2026-09-29

Source Vals AI

Page 1 of 2

Evidence

Sources and data date

Every statement links to its underlying documentation or leaderboard.

Type	Evidence and data date
Input	Documented modalities. Retrieved September 8, 2026. Source: Model documentation · Input (Model documentation)
Output	Documented modalities. Retrieved September 8, 2026. Source: Model documentation · Input (Model documentation)
Research status	Research date September 8, 2026. 8 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not found in the inspected sources	knowledge-cutoff Original wording The checked original card, launch post and model overview do not publish an exact cutoff for command-r-03-2024.
Not found in the inspected sources	training-tokens Original wording The checked original sources do not publish a training-token count.
Unresolved	refresh-model-exclusion Original wording Earlier evidence used the same canonical ID for the August 2024 refresh. The refresh is a separate command-r-08-2024 model and its facts and benchmarks are intentionally excluded from this original record.
Unresolved	AA benchmark rows Original wording No exact Artificial Analysis model page located in the checked public sitemap.
Additional source	Context window data (retrieved September 8, 2026) · Cohere Command R model card · Editorial description reviewed October 2, 2026
Additional source	Provider model lifecycle (retrieved October 2, 2026) · Cohere model overview and lifecycle · Editorial description reviewed October 2, 2026
Additional source	Vals AI (retrieved August 11, 2026; September 29, 2026)
Additional source	Open LLM leaderboard table in Cohere Command R+ card (retrieved September 8, 2026)
Additional source	Model metadata (retrieved September 8, 2026) · Cohere Command R announcement · Editorial description reviewed October 2, 2026
Additional source	Model metadata (retrieved September 8, 2026)
Additional source	Model metadata (retrieved September 8, 2026)
Additional source	OpenRouter: Popularity index (derived), CC BY 4.0
Additional source	Vercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional source	Creative Commons Attribution 4.0

Specifications and access

Published benchmarks

Legal

LegalBench

LegalBench

Measurements without matching peer values

Recorded measurement details

Popularity

Sources and data date