Mistral AILanguage model
Codestral (latest)
Codestral (latest) has no overall index value yet because there are too few independent benchmark results.
16out of 100same as last month
based on two data groups, mostly search
Summary
Codestral (latest) is Mistral’s rolling API entry for code completion and fill-in-the-middle generation. The alias bundles the current service, while Codestral-22B-v0.1 identifies a fixed checkpoint.
Mistral’s documentation lists additional dated Codestral versions. The alias can change without changing older model cards or locally loaded weights.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 6 published benchmark results
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.
JMigBench (Show measurement, test conditions, and source)
- Source value
- 0.63
- Score
- 0.63
- Metric
- CodeBLEU similarity, Generated Java 11 vs True Java 11
- Unit
- ratio
- Direction
- Higher is better
- Sample
- 40
- Source type
- independent-evaluator
- Evaluator
- Nishil Amin et al., JMigBench, ACM ReCode ’26 (arXiv:2602.09930)
- Status
- archived
- Retrieved at
- 2026-10-18
- Methodology
- Independent evaluation. Mistral API route `codestral-latest`; the authors state that during their September 2025 experiments this alias referred to Codestral 2501. The model received one standardized role-based Java 8-to-11 migration prompt per function, with no iterative refinement, on the revised 45-function-pair dataset. API decoding parameters and tool settings are not stated. This is a historical route observation, not evidence for the mutable alias's current resolution.
- Context
- Archived historical observation only. The authors report that the rolling Mistral API route "codestral-latest" resolved to Codestral 2501 during their September 2025 experiment. This does not establish today’s backend or today’s performance for the mutable alias. Table 2 averages 40 of 45 functions because five lacked Dataflow Match scores; the authors state that generated Java was not compiled or executed.
JMigBench (Show measurement, test conditions, and source)
- Source value
- 0.57
- Score
- 0.57
- Metric
- N-gram Match, Generated Java 11 vs True Java 11
- Unit
- ratio
- Direction
- Higher is better
- Sample
- 40
- Source type
- independent-evaluator
- Evaluator
- Nishil Amin et al., JMigBench, ACM ReCode ’26 (arXiv:2602.09930)
- Status
- archived
- Retrieved at
- 2026-10-18
- Methodology
- Independent evaluation. Mistral API route `codestral-latest`; the authors state that during their September 2025 experiments this alias referred to Codestral 2501. The model received one standardized role-based Java 8-to-11 migration prompt per function, with no iterative refinement, on the revised 45-function-pair dataset. API decoding parameters and tool settings are not stated. This is a historical route observation, not evidence for the mutable alias's current resolution.
- Context
- Archived historical observation only. The authors report that the rolling Mistral API route "codestral-latest" resolved to Codestral 2501 during their September 2025 experiment. This does not establish today’s backend or today’s performance for the mutable alias. Table 2 averages 40 of 45 functions because five lacked Dataflow Match scores; the authors state that generated Java was not compiled or executed.
JMigBench (Show measurement, test conditions, and source)
- Source value
- 0.66
- Score
- 0.66
- Metric
- Weighted N-gram Match, Generated Java 11 vs True Java 11
- Unit
- ratio
- Direction
- Higher is better
- Sample
- 40
- Source type
- independent-evaluator
- Evaluator
- Nishil Amin et al., JMigBench, ACM ReCode ’26 (arXiv:2602.09930)
- Status
- archived
- Retrieved at
- 2026-10-18
- Methodology
- Independent evaluation. Mistral API route `codestral-latest`; the authors state that during their September 2025 experiments this alias referred to Codestral 2501. The model received one standardized role-based Java 8-to-11 migration prompt per function, with no iterative refinement, on the revised 45-function-pair dataset. API decoding parameters and tool settings are not stated. This is a historical route observation, not evidence for the mutable alias's current resolution.
- Context
- Archived historical observation only. The authors report that the rolling Mistral API route "codestral-latest" resolved to Codestral 2501 during their September 2025 experiment. This does not establish today’s backend or today’s performance for the mutable alias. Table 2 averages 40 of 45 functions because five lacked Dataflow Match scores; the authors state that generated Java was not compiled or executed.
JMigBench (Show measurement, test conditions, and source)
- Source value
- 0.68
- Score
- 0.68
- Metric
- Syntax Match, Generated Java 11 vs True Java 11
- Unit
- ratio
- Direction
- Higher is better
- Sample
- 40
- Source type
- independent-evaluator
- Evaluator
- Nishil Amin et al., JMigBench, ACM ReCode ’26 (arXiv:2602.09930)
- Status
- archived
- Retrieved at
- 2026-10-18
- Methodology
- Independent evaluation. Mistral API route `codestral-latest`; the authors state that during their September 2025 experiments this alias referred to Codestral 2501. The model received one standardized role-based Java 8-to-11 migration prompt per function, with no iterative refinement, on the revised 45-function-pair dataset. API decoding parameters and tool settings are not stated. This is a historical route observation, not evidence for the mutable alias's current resolution.
- Context
- Archived historical observation only. The authors report that the rolling Mistral API route "codestral-latest" resolved to Codestral 2501 during their September 2025 experiment. This does not establish today’s backend or today’s performance for the mutable alias. Table 2 averages 40 of 45 functions because five lacked Dataflow Match scores; the authors state that generated Java was not compiled or executed.
JMigBench (Show measurement, test conditions, and source)
- Source value
- 0.62
- Score
- 0.62
- Metric
- Dataflow Match, Generated Java 11 vs True Java 11
- Unit
- ratio
- Direction
- Higher is better
- Sample
- 40
- Source type
- independent-evaluator
- Evaluator
- Nishil Amin et al., JMigBench, ACM ReCode ’26 (arXiv:2602.09930)
- Status
- archived
- Retrieved at
- 2026-10-18
- Methodology
- Independent evaluation. Mistral API route `codestral-latest`; the authors state that during their September 2025 experiments this alias referred to Codestral 2501. The model received one standardized role-based Java 8-to-11 migration prompt per function, with no iterative refinement, on the revised 45-function-pair dataset. API decoding parameters and tool settings are not stated. This is a historical route observation, not evidence for the mutable alias's current resolution.
- Context
- Archived historical observation only. The authors report that the rolling Mistral API route "codestral-latest" resolved to Codestral 2501 during their September 2025 experiment. This does not establish today’s backend or today’s performance for the mutable alias. Table 2 averages 40 of 45 functions because five lacked Dataflow Match scores; the authors state that generated Java was not compiled or executed.
JMigBench (Show measurement, test conditions, and source)
- Source value
- 5
- Score
- 5
- Metric
- Complete Matches, functions identical lexically and semantically to the true Java 11 function
- Unit
- count
- Direction
- Higher is better
- Sample
- 45
- Source type
- independent-evaluator
- Evaluator
- Nishil Amin et al., JMigBench, ACM ReCode ’26 (arXiv:2602.09930)
- Status
- archived
- Retrieved at
- 2026-10-18
- Methodology
- Independent evaluation. Mistral API route `codestral-latest`; the authors state that during their September 2025 experiments this alias referred to Codestral 2501. The model received one standardized role-based Java 8-to-11 migration prompt per function, with no iterative refinement, on the revised 45-function-pair dataset. API decoding parameters and tool settings are not stated. This is a historical route observation, not evidence for the mutable alias's current resolution.
- Context
- Archived historical observation only. The authors report that the rolling Mistral API route "codestral-latest" resolved to Codestral 2501 during their September 2025 experiment. This does not establish today’s backend or today’s performance for the mutable alias. Table 2 averages 40 of 45 functions because five lacked Dataflow Match scores; the authors state that generated Java was not compiled or executed.
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
Codestral (latest)Muse Wonder Full 16Ornith 1.0 31B 16Solar Mini 16Code Llama 70B Instruct 16
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 Codestral (latest), chosen by model line, developer, size, price and release date.
- Codestral-22B-v0.1Mistral AI · same developer
- Magistral SmallMistral AI · same developer
- Devstral SmallMistral AI · same developer
- Pixtral 12BMistral AI · same developer
- Magistral Small 1.2Mistral AI · same developer
- Devstral Small 2Mistral AI · same developer
- Mixtral 8x7BMistral AI · same developer
- Mistral 7B v0.3Mistral AI · 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 Codestral (latest)
What is the knowledge cutoff of Codestral (latest)?
October 2024, according to Mistral AI.
Stated by Mistral AI
How large is the context window of Codestral (latest)?
256,000 tokens. That is how much text Codestral (latest) reads at once, about 380 A4 pages.
Page count as a rule of thumb from Artificial Analysis
Can I run Codestral (latest) locally?
Yes, the weights are open.
Stated by Mistral AI
- Open weights
- yes
- License
- Mistral Non-Production License
- Parameters
- 22B
- active per token
- n/a
Sources and data date
Show all 16 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. 9 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Unresolved | current-alias-checkpoint Note No current public primary-source bridge from the moving Codestral alias to an exact benchmarked checkpoint was established. The historical 22B and 25.01 rows are not copied. |
| Unresolved | dated-alias-checkpoint-benchmarks Note Third-party pages display scores for Codestral latest, but the inspected material does not bind each score to an individually sourced run or establish which dated checkpoint the mutable alias served. Mistral explicitly warns that latest aliases can update silently. No fixed codestral-2508 score is assigned to this alias. |
| Unresolved | quality-benchmarks Note No numeric benchmark with a dated resolution of the mutable Codestral latest alias was found. Fixed-version Codestral results and third-party scores without an exact alias resolution are not assigned to this profile. |
| Additional source | Hugging Face (retrieved October 15, 2026) · Codestral 22B v0.1 model card · Editorial description reviewed October 2, 2026 |
| Additional source | Nishil Amin et al., JMigBench, ACM ReCode ’26 (arXiv:2602.09930) (retrieved October 18, 2026) |
| Additional source | Model metadata (retrieved October 16, 2026) · Codestral model documentation · Editorial description reviewed October 2, 2026 |
| Additional source | Model metadata (retrieved October 16, 2026)12 (unavailable on November 2, 2026) https://openrouter.ai/mistralai/codestral-latest3 |
| Additional source | Model metadata (retrieved October 18, 2026)12 |
| Additional source | Codestral announcement · 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 |