CohereLanguage model
Command R7B Arabic
Command R7B Arabic has no overall index value yet because there are too few independent benchmark results.
27out of 100-2 vs. last month
based on search, usage and attention
Summary
Command R7B Arabic is Cohere's Arabic-focused R7B variant for Modern Standard Arabic and English. Cohere highlights instruction following, length control, RAG, and reduced unwanted code-switching. The model card also emphasizes knowledge and understanding of Arabic language and cultures. That makes this checkpoint more specialized than general Command R7B.
The open 8-billion-parameter model provides 128,000 context tokens, tool use, and a 4,000-token output limit. Cohere documents separate conversational and instruct modes plus a RAG chat template with citations. CC-BY-NC-4.0 limits commercial use, while the checkpoint remains specialized for Arabic and English.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 66 published benchmark results
Page 1 of 6
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.
Command R7B Arabic BBH (Show measurement, test conditions, and source)
- Source value
- 36.2
- Score
- 36.2
- Metric
- accuracy or task score as published
- Unit
- %
- Benchmark version
- Standardized Hugging Face Open LLM Leaderboard result in Table 2
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere, Command R7B Arabic paper, AfricanNLP 2025
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Exact Command R7B Arabic row in Table 2. The paper reports percentage performance but no task-specific prompts, decoding details, or scored-item denominator.
- Context
- Table 2 does not report a per-benchmark sample count or task-specific evaluation setup.
Command R7B Arabic MuSR (Show measurement, test conditions, and source)
- Source value
- 11.9
- Score
- 11.9
- Metric
- accuracy or task score as published
- Unit
- %
- Benchmark version
- Standardized Hugging Face Open LLM Leaderboard result in Table 2
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere, Command R7B Arabic paper, AfricanNLP 2025
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Exact Command R7B Arabic row in Table 2. The paper reports percentage performance but no task-specific prompts, decoding details, or scored-item denominator.
- Context
- Table 2 does not report a per-benchmark sample count or task-specific evaluation setup.
Command R7B Arabic GPQA (Show measurement, test conditions, and source)
- Source value
- 7.9
- Score
- 7.9
- Metric
- accuracy or task score as published
- Unit
- %
- Benchmark version
- Standardized Hugging Face Open LLM Leaderboard result in Table 2
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere, Command R7B Arabic paper, AfricanNLP 2025
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Exact Command R7B Arabic row in Table 2. The paper reports percentage performance but no task-specific prompts, decoding details, or scored-item denominator.
- Context
- Table 2 does not report a per-benchmark sample count or task-specific evaluation setup.
Command R7B Arabic MMLU-Pro (Show measurement, test conditions, and source)
- Source value
- 29.4
- Score
- 29.4
- Metric
- accuracy or task score as published
- Unit
- %
- Benchmark version
- Standardized Hugging Face Open LLM Leaderboard result in Table 2
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere, Command R7B Arabic paper, AfricanNLP 2025
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Exact Command R7B Arabic row in Table 2. The paper reports percentage performance but no task-specific prompts, decoding details, or scored-item denominator.
- Context
- Table 2 does not report a per-benchmark sample count or task-specific evaluation setup.
Command R7B Arabic IFEval (Show measurement, test conditions, and source)
- Source value
- 83.3
- Score
- 83.3
- Metric
- accuracy or task score as published
- Unit
- %
- Benchmark version
- Standardized Hugging Face Open LLM Leaderboard result in Table 2
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere, Command R7B Arabic paper, AfricanNLP 2025
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Exact Command R7B Arabic row in Table 2. The paper reports percentage performance but no task-specific prompts, decoding details, or scored-item denominator.
- Context
- Table 2 does not report a per-benchmark sample count or task-specific evaluation setup.
Command R7B Arabic MATH (Show measurement, test conditions, and source)
- Source value
- 19.6
- Score
- 19.6
- Metric
- accuracy or task score as published
- Unit
- %
- Benchmark version
- Standardized Hugging Face Open LLM Leaderboard result, Table 2; MATH split changed in early January 2025
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Cohere, Command R7B Arabic paper, AfricanNLP 2025
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Exact Command R7B Arabic row in Table 2. The paper reports percentage performance but no task-specific prompts, decoding details, or scored-item denominator. The authors state that MATH split changed in early January 2025 after a DMCA takedown.
- Context
- The benchmark split changed after a DMCA notice; the replacement split and denominator are not identified.
Hala technical report, Arabic MMLU (Show measurement, test conditions, and source)
- Source value
- 59.3
- Score
- 59.3
- Metric
- accuracy
- Unit
- %
- Benchmark version
- Hala Technical Report 2026, Table 1, OALL task suite excluding Alrage
- Category
- multilingual
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Hala Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact Table 1 checkpoint CohereForAI/c4ai-command-r7b-arabic-02-2025 (7B); Arabic MMLU score from the authors' Arabic-centric evaluation. The report says all rows use LightEval with vLLM; it does not give a row-specific task revision, generation settings, or evaluated N here.
- Context
- Task scores come from a different evaluation setup than similarly named Command R7B Arabic profile observations. No average or per-task denominator is inferred.
Hala technical report, AlGhafa (Show measurement, test conditions, and source)
- Source value
- 74.8
- Score
- 74.8
- Metric
- accuracy
- Unit
- %
- Benchmark version
- Hala Technical Report 2026, Table 1, OALL task suite excluding Alrage
- Category
- multilingual
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Hala Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact Table 1 checkpoint CohereForAI/c4ai-command-r7b-arabic-02-2025 (7B); AlGhafa score from the authors' Arabic-centric evaluation. The report says all rows use LightEval with vLLM; it does not give a row-specific task revision, generation settings, or evaluated N here.
- Context
- Task scores come from a different evaluation setup than similarly named Command R7B Arabic profile observations. No average or per-task denominator is inferred.
Hala technical report, EXAMS (Show measurement, test conditions, and source)
- Source value
- 65
- Score
- 65
- Metric
- accuracy
- Unit
- %
- Benchmark version
- Hala Technical Report 2026, Table 1, OALL task suite excluding Alrage
- Category
- multilingual
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Hala Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact Table 1 checkpoint CohereForAI/c4ai-command-r7b-arabic-02-2025 (7B); EXAMS score from the authors' Arabic-centric evaluation. The report says all rows use LightEval with vLLM; it does not give a row-specific task revision, generation settings, or evaluated N here.
- Context
- Task scores come from a different evaluation setup than similarly named Command R7B Arabic profile observations. No average or per-task denominator is inferred.
Hala technical report, MadinahQA (Show measurement, test conditions, and source)
- Source value
- 63.8
- Score
- 63.8
- Metric
- accuracy
- Unit
- %
- Benchmark version
- Hala Technical Report 2026, Table 1, OALL task suite excluding Alrage
- Category
- multilingual
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Hala Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact Table 1 checkpoint CohereForAI/c4ai-command-r7b-arabic-02-2025 (7B); MadinahQA score from the authors' Arabic-centric evaluation. The report says all rows use LightEval with vLLM; it does not give a row-specific task revision, generation settings, or evaluated N here.
- Context
- Task scores come from a different evaluation setup than similarly named Command R7B Arabic profile observations. No average or per-task denominator is inferred.
Hala technical report, AraTrust (Show measurement, test conditions, and source)
- Source value
- 80.5
- Score
- 80.5
- Metric
- accuracy
- Unit
- %
- Benchmark version
- Hala Technical Report 2026, Table 1, OALL task suite excluding Alrage
- Category
- multilingual
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Hala Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact Table 1 checkpoint CohereForAI/c4ai-command-r7b-arabic-02-2025 (7B); AraTrust score from the authors' Arabic-centric evaluation. The report says all rows use LightEval with vLLM; it does not give a row-specific task revision, generation settings, or evaluated N here.
- Context
- Task scores come from a different evaluation setup than similarly named Command R7B Arabic profile observations. No average or per-task denominator is inferred.
Hala technical report, ArbMMLU-HT (Show measurement, test conditions, and source)
- Source value
- 50.1
- Score
- 50.1
- Metric
- accuracy
- Unit
- %
- Benchmark version
- Hala Technical Report 2026, Table 1, OALL task suite excluding Alrage
- Category
- multilingual
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Hala Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact Table 1 checkpoint CohereForAI/c4ai-command-r7b-arabic-02-2025 (7B); ArbMMLU-HT score from the authors' Arabic-centric evaluation. The report says all rows use LightEval with vLLM; it does not give a row-specific task revision, generation settings, or evaluated N here.
- Context
- Task scores come from a different evaluation setup than similarly named Command R7B Arabic profile observations. No average or per-task denominator is inferred.
Page 1 of 6
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 R7B ArabicSolar Pro 2 28Command A Vision 27Grok 4.5 27Claude Opus 4.7 27
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 R7B Arabic, chosen by model line, developer, size, price and release date.
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 R7B Arabic
What is the knowledge cutoff of Command R7B Arabic?
June 2024, according to Cohere. That is 8 months before release.
Stated by Cohere
How large is the context window of Command R7B Arabic?
128,000 tokens. That is how much text Command R7B Arabic reads at once, about 190 A4 pages.
Page count as a rule of thumb from Artificial Analysis
Can I run Command R7B Arabic locally?
Yes, the weights are open.
Stated by Cohere
- Open weights
- yes
- License
- CC BY-NC-4.0
- Parameters
- 8B
- active per token
- n/a
Sources and data date
Show all 15 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.
Without vendor claims and historical or disputed results; those stay labelled in the single values.
| Type | Evidence and data date |
|---|---|
| Input | Documented modalities. Retrieved October 27, 2026. |
| Output | Documented modalities. Retrieved October 27, 2026. |
| Research status | Research date October 15, 2026. 6 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Not published | evaluation-details Note The model card does not identify all dataset versions, prompts, decoding settings, score units or evaluation dates. |
| Not published | MedAraBench model-specific sample count Original wording The paper reports 4,989 test items for the whole dataset but not this model's successful-response denominator or per-table-cell N; no sampleCount is entered. |
| Not published | MedAraBench confidence intervals Original wording No confidence intervals for the exact model's overall, specialty, or difficulty-level results are published. |
| Not published | MedAraBench table-cell result validity Original wording The paper reports raw accuracy ratios but does not publish per-cell completion/parse-failure counts; the included values are exactly the table entries, without inferred denominators. |
| Additional source | Hugging Face · Cohere Command R7B Arabic model card · Editorial description reviewed October 2, 2026 |
| Additional source | Cohere Command R7B Arabic model-card multilingual evaluation table (retrieved October 15, 2026; October 17, 2026) |
| Additional source | Cohere, Command R7B Arabic paper, AfricanNLP 2025 (retrieved October 17, 2026) |
| Additional source | MedAraBench paper, arXiv:2602.01714v1 (retrieved October 17, 2026) |
| Additional source | Alexandria: Context-Aware Machine Translation for Arabic Dialects (retrieved October 17, 2026) |
| Additional source | Current state of LLMs for Arabic dialectal machine translation (retrieved October 17, 2026) |
| Additional source | Hala Technical Report (retrieved October 17, 2026) |
| Additional source | Cohere Command R7B Arabic release note · Editorial description reviewed October 2, 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 | Ollama: cumulative downloads for the popularity index |
| Additional source | Replicate: cumulative runs for the popularity index |
| Additional source | Google Trending Now: limited feed selection for Germany and the US |
| Additional source | Creative Commons Attribution 4.0 |