diff --git a/docs/capabilities/batch.md b/docs/capabilities/batch.md index 570158a..95ff53f 100644 --- a/docs/capabilities/batch.md +++ b/docs/capabilities/batch.md @@ -405,10 +405,9 @@ Currently, there is a maximum limit of 1 million pending requests per workspace. Currently, there is no maximum limit. ### How long does the batch API take to process? -Batch generation can take up to 24 hours, although it might finish earlier. Please note that processing speeds may be adjusted based on current demand and the volume of your request. Your batch results will only be accessible once the entire batch processing is complete. - -Users can set `timeout_hours` when creating a job, which specifies the number of hours after which the job should expire. It defaults to 24 hours and cannot exceed this limit at present. This means that batches will expire if processing does not complete within 24 hours. This is subject to change in the future. +Processing speeds may be adjusted based on current demand and the volume of your request. Your batch results will only be accessible once the entire batch processing is complete. +Users can set `timeout_hours` when creating a job, which specifies the number of hours after which the job should expire. It defaults to 24 hours and should be lower than 7 days. A batch will expire if processing does not complete within the specified timeout. ### Can I view batch results from my workspace? Yes, batches are specific to a workspace. You can see all batches and their results that were created within the workspace associated with your API key. diff --git a/docs/capabilities/finetuning.mdx b/docs/capabilities/finetuning.mdx index caf07d3..0325217 100644 --- a/docs/capabilities/finetuning.mdx +++ b/docs/capabilities/finetuning.mdx @@ -238,7 +238,7 @@ curl https://api.mistral.ai/v1/files \ ## Create a fine-tuning job The next step is to create a fine-tuning job. -- model: the specific model you would like to fine-tune. The choices are `open-mistral-7b` (v0.3), `mistral-small-latest` (`mistral-small-2409`), `codestral-latest` (`codestral-2405`), `open-mistral-nemo` and , `mistral-large-latest` (`mistral-large-2407`). +- model: the specific model you would like to fine-tune. The choices are `open-mistral-7b` (v0.3), `mistral-small-latest` (`mistral-small-2409`), `codestral-latest` (`codestral-2405`), `open-mistral-nemo`, `mistral-large-latest` (`mistral-large-2411`), and `ministral-8b-latest`(`ministral-3b-2410`). - training_files: a collection of training file IDs, which can consist of a single file or multiple files - validation_files: a collection of validation file IDs, which can consist of a single file or multiple files - hyperparameters: two adjustable hyperparameters, "training_step" and "learning_rate", that users can modify. @@ -272,7 +272,7 @@ created_jobs ```typescript -const createdJob = await client.fineTuning.jobs.create({jobIn:{ +const createdJob = await client.fineTuning.jobs.create({ model: 'open-mistral-7b', trainingFiles: [{fileId: training_data.id, weight: 1}], validationFiles: [validation_data.id], @@ -281,7 +281,7 @@ const createdJob = await client.fineTuning.jobs.create({jobIn:{ learningRate: 0.0001, }, autoStart:false, - }}); + }); await client.fineTuning.jobs.start({jobId: createdJob.id}) ``` diff --git a/docs/deployment/cloud/aws.mdx b/docs/deployment/cloud/aws.mdx index 98431e4..eb66172 100644 --- a/docs/deployment/cloud/aws.mdx +++ b/docs/deployment/cloud/aws.mdx @@ -15,10 +15,10 @@ to manage any infrastructure. As of today, the following models are available: -- Mistral 7B +- Mistral Large (24.07, 24.02) +- Mistral Small (24.02) - Mixtral 8x7B -- Mistral Small -- Mistral Large +- Mistral 7B For more details, visit the [models](../../../getting-started/models/models_overview/) page. diff --git a/docs/deployment/cloud/azure.mdx b/docs/deployment/cloud/azure.mdx index 3b7c1fd..aaf6d57 100644 --- a/docs/deployment/cloud/azure.mdx +++ b/docs/deployment/cloud/azure.mdx @@ -21,9 +21,10 @@ in two ways: This page focuses on the MaaS offering, where the following models are available: -- Mistral Large -- Mistral Small -- Mistral NeMo +- Mistral Large (24.11, 24.07) +- Mistral Small (24.09) +- Ministral 3B (24.10) +- Mistral Nemo For more details, visit the [models](../../../getting-started/models/models_overview) page. diff --git a/docs/deployment/cloud/outscale.mdx b/docs/deployment/cloud/outscale.mdx index b6ddd25..645e5ea 100644 --- a/docs/deployment/cloud/outscale.mdx +++ b/docs/deployment/cloud/outscale.mdx @@ -16,8 +16,9 @@ on your behalf, provision a virtual machine and a GPU then deploy the model on i As of today, the following models are available: -- Mistral Small (2409) -- Codestral +- Mistral Small (24.09) +- Codestral (24.05) +- Ministral 8B (24.10) For more details, visit the [models](../../../getting-started/models/models_overview) page. diff --git a/docs/deployment/cloud/vertex.mdx b/docs/deployment/cloud/vertex.mdx index e619f8f..0fe2027 100644 --- a/docs/deployment/cloud/vertex.mdx +++ b/docs/deployment/cloud/vertex.mdx @@ -15,9 +15,9 @@ so you don't have to manage any infrastructure. As of today, the following models are available: -- Mistral Large -- Mistral NeMo -- Codestral (chat and FIM completions) +- Mistral Large (24.11, 24.07) +- Codestral (24.05) +- Mistral Nemo For more details, visit the [models](../../../getting-started/models/models_overview) page. diff --git a/docs/getting-started/changelog.mdx b/docs/getting-started/changelog.mdx index c895297..b600743 100644 --- a/docs/getting-started/changelog.mdx +++ b/docs/getting-started/changelog.mdx @@ -3,6 +3,8 @@ id: changelog title: Changelog sidebar_position: 1.8 --- +January 13, 2025 +- We released Codestral 25.01 (`codestral-2501`). November 18, 2024 - We released Mistral Large 24.11 (`mistral-large-2411`) and Pixtral Large (`pixtral-large-2411`). diff --git a/docs/getting-started/introduction.mdx b/docs/getting-started/introduction.mdx index de30af8..0e2ac53 100644 --- a/docs/getting-started/introduction.mdx +++ b/docs/getting-started/introduction.mdx @@ -16,20 +16,18 @@ We release both premier models and free models, driving innovation and convenien - Mistral Large, our top-tier reasoning model for high-complexity tasks with the lastest version released [November 2024](https://mistral.ai/news/pixtral-large/) - Pixtral Large, our frontier-class multimodal model released [November 2024](https://mistral.ai/news/pixtral-large/) -- Ministral 3B, world’s best edge model released [October 2024](https://mistral.ai/news/ministraux/). -- Ministral 8B, powerful edge model with extremely high performance/price ratio released [October 2024](https://mistral.ai/news/ministraux/). -- Mistral Small, our latest enterprise-grade small model with the lastest version v2 released [September 2024](https://mistral.ai/news/september-24-release/). -- Codestral, our cutting-edge language model for coding released [May 2024](https://mistral.ai/news/codestral/) +- Ministral 3B, world’s best edge model released [October 2024](https://mistral.ai/news/ministraux/) +- Ministral 8B, powerful edge model with extremely high performance/price ratio released [October 2024](https://mistral.ai/news/ministraux/) +- Mistral Small, our latest enterprise-grade small model with the lastest version v2 released [September 2024](https://mistral.ai/news/september-24-release/) +- Codestral, our cutting-edge language model for coding with the latest version released [January 2025](https://mistral.ai/news/codestral-2501/) - Mistral Embed, our state-of-the-art semantic for extracting representation of text extracts +- Mistral Moderation, our moderation service that enables our users to detect harmful text content ### Free models - Pixtral, a 12B model with image understanding capabilities in addition to text released [September 2024](https://mistral.ai/news/pixtral-12b/) - Mistral NeMo, our best multilingual open source model released [July 2024](https://mistral.ai/news/mistral-nemo/) - Codestral Mamba, our first mamba 2 open source model released [July 2024](https://mistral.ai/news/codestral-mamba/) - Mathstral 7b, our first math open source model released [July 2024](https://mistral.ai/news/mathstral/) -- Mixtral 8x22b, our best open source model to date released [April 2024](https://mistral.ai/news/mixtral-8x22b/) -- Mixtral 8x7b, our first sparse mixture-of-experts released [December 2023](https://mistral.ai/news/mixtral-of-experts/) -- Mistral 7b, our first dense model released [September 2023](https://mistral.ai/news/announcing-mistral-7b/) Learn more about our models [here](/getting-started/models/models_overview). diff --git a/docs/getting-started/models/benchmark.md b/docs/getting-started/models/benchmark.md index 2cafbf1..5b66729 100644 --- a/docs/getting-started/models/benchmark.md +++ b/docs/getting-started/models/benchmark.md @@ -15,7 +15,7 @@ You can find the benchmark results in the following blog posts: - [Mistral Large](https://mistral.ai/news/mistral-large-2407/): a cutting-edge text generation model with top-tier reasoning capabilities. It can be used for complex multilingual reasoning tasks, including text understanding, transformation, and code generation. - [Mistral Nemo](https://mistral.ai/news/mistral-nemo/): Mistral Nemo's reasoning, world knowledge, and coding performance are state-of-the-art in its size category. As it relies on standard architecture, Mistral Nemo is easy to use and a drop-in replacement in any system using Mistral 7B that it supersedes. -- [Codestral](https://mistral.ai/news/codestral/): as a 22B model, Codestral sets a new standard on the performance/latency space for code generation compared to previous models used for coding. +- [Codestral](https://mistral.ai/news/codestral-2501/): Codestral sets a new standard on the performance/latency space for code generation compared to previous models used for coding. - [Codestral-Mamba](https://mistral.ai/news/codestral-mamba/): we have trained this model with advanced code and reasoning capabilities, enabling the model to have a strong performance on par with SOTA transformer-based models. - [Mathstral](https://mistral.ai/news/mathstral/): Mathstral stands on the shoulders of Mistral 7B and specialises in STEM subjects. It achieves state-of-the-art reasoning capacities in its size category across various industry-standard benchmarks. - [Mixtral 8x22B](https://mistral.ai/news/mixtral-8x22b/): our most performant open model. It handles English, diff --git a/docs/getting-started/models/overview.md b/docs/getting-started/models/overview.md index 4214c07..f829957 100644 --- a/docs/getting-started/models/overview.md +++ b/docs/getting-started/models/overview.md @@ -16,12 +16,12 @@ Mistral provides two types of models: free models and premier models. | Model | Weight availability|Available via API| Description | Max Tokens| API Endpoints|Version| |--------------------|:--------------------:|:--------------------:|:--------------------:|:--------------------:|:--------------------:|:--------------------:| -| Mistral Large |:heavy_check_mark:
[Mistral Research License](https://mistral.ai/licenses/MRL-0.1.md)| :heavy_check_mark: |Our top-tier reasoning model for high-complexity tasks with the lastest version released November 2024. Learn more on our [blog post](https://mistral.ai/news/pixtral-large/) | 128k | `mistral-large-latest`| 24.11| -| Pixtral Large |:heavy_check_mark:
[Mistral Research License](https://mistral.ai/licenses/MRL-0.1.md)| :heavy_check_mark: |Our frontier-class multimodal model released November 2024. Learn more on our [blog post](https://mistral.ai/news/pixtral-large/)| 128k | `pixtral-large-latest`| 24.11| -| Ministral 3B | | :heavy_check_mark: | World’s best edge model. Learn more on our [blog post](https://mistral.ai/news/ministraux/) | 128k | `ministral-3b-latest` | 24.10| -| Ministral 8B | :heavy_check_mark:
[Mistral Research License](https://mistral.ai/licenses/MRL-0.1.md)| :heavy_check_mark: |Powerful edge model with extremely high performance/price ratio. Learn more on our [blog post](https://mistral.ai/news/ministraux/) | 128k | `ministral-8b-latest` | 24.10| +| Codestral | | :heavy_check_mark: | Our cutting-edge language model for coding with the second version released January 2025, Codestral specializes in low-latency, high-frequency tasks such as fill-in-the-middle (FIM), code correction and test generation. Learn more on our blog post(https://mistral.ai/news/codestral-2501/) | 256k | `codestral-latest` | 25.01| +| Mistral Large |:heavy_check_mark:
[Mistral Research License](https://mistral.ai/licenses/MRL-0.1.md)| :heavy_check_mark: |Our top-tier reasoning model for high-complexity tasks with the lastest version released November 2024. Learn more on our [blog post](https://mistral.ai/news/pixtral-large/) | 131k | `mistral-large-latest`| 24.11| +| Pixtral Large |:heavy_check_mark:
[Mistral Research License](https://mistral.ai/licenses/MRL-0.1.md)| :heavy_check_mark: |Our frontier-class multimodal model released November 2024. Learn more on our [blog post](https://mistral.ai/news/pixtral-large/)| 131k | `pixtral-large-latest`| 24.11| +| Ministral 3B | | :heavy_check_mark: | World’s best edge model. Learn more on our [blog post](https://mistral.ai/news/ministraux/) | 131k | `ministral-3b-latest` | 24.10| +| Ministral 8B | :heavy_check_mark:
[Mistral Research License](https://mistral.ai/licenses/MRL-0.1.md)| :heavy_check_mark: |Powerful edge model with extremely high performance/price ratio. Learn more on our [blog post](https://mistral.ai/news/ministraux/) | 131k | `ministral-8b-latest` | 24.10| | Mistral Small | :heavy_check_mark:
[Mistral Research License](https://mistral.ai/licenses/MRL-0.1.md) | :heavy_check_mark: | Our latest enterprise-grade small model with the lastest version v2 released September 2024. Learn more on our [blog post](https://mistral.ai/news/september-24-release/) | 32k | `mistral-small-latest` | 24.09| -| Codestral |:heavy_check_mark:
[Mistral Non-Production License](https://mistral.ai/licenses/MNPL-0.1.md) | :heavy_check_mark: | Our cutting-edge language model for coding released May 2024 | 32k | `codestral-latest` | 24.05| | Mistral Embed | | :heavy_check_mark: | Our state-of-the-art semantic for extracting representation of text extracts | 8k | `mistral-embed` | 23.12| | Mistral Moderation | | :heavy_check_mark: | Our moderation service that enables our users to detect harmful text content | 8k | `mistral-moderation-latest` | 24.11| @@ -32,13 +32,13 @@ Mistral provides two types of models: free models and premier models. | Model | Weight availability|Available via API| Description | Max Tokens| API Endpoints|Version| |--------------------|:--------------------:|:--------------------:|:--------------------:|:--------------------:|:--------------------:|:--------------------:| -| Pixtral | :heavy_check_mark:
Apache2 | :heavy_check_mark: | A 12B model with image understanding capabilities in addition to text. Learn more on our [blog post](https://mistral.ai/news/pixtral-12b/)| 128k | `pixtral-12b-2409` | 24.09| +| Pixtral | :heavy_check_mark:
Apache2 | :heavy_check_mark: | A 12B model with image understanding capabilities in addition to text. Learn more on our [blog post](https://mistral.ai/news/pixtral-12b/)| 131k | `pixtral-12b-2409` | 24.09| - **Research models** | Model | Weight availability|Available via API| Description | Max Tokens| API Endpoints|Version| |--------------------|:--------------------:|:--------------------:|:--------------------:|:--------------------:|:--------------------:|:--------------------:| -| Mistral Nemo | :heavy_check_mark:
Apache2 | :heavy_check_mark: | Our best multilingual open source model released July 2024. Learn more on our [blog post](https://mistral.ai/news/mistral-nemo/) | 128k | `open-mistral-nemo`| 24.07| +| Mistral Nemo | :heavy_check_mark:
Apache2 | :heavy_check_mark: | Our best multilingual open source model released July 2024. Learn more on our [blog post](https://mistral.ai/news/mistral-nemo/) | 131k | `open-mistral-nemo`| 24.07| | Codestral Mamba | :heavy_check_mark:
Apache2 | :heavy_check_mark: | Our first mamba 2 open source model released July 2024. Learn more on our [blog post](https://mistral.ai/news/codestral-mamba/) | 256k | `open-codestral-mamba`| v0.1| | Mathstral | :heavy_check_mark:
Apache2 | | Our first math open source model released July 2024. Learn more on our [blog post](https://mistral.ai/news/mathstral/) | 32k | NA| v0.1| @@ -58,7 +58,7 @@ Here are the details of the available versions: - `ministral-8b-latest`: currently points to `ministral-8b-2410`. - `open-mistral-nemo`: currently points to `open-mistral-nemo-2407`. - `mistral-small-latest`: currently points to `mistral-small-2409`. -- `codestral-latest`: currently points to `codestral-2405`. +- `codestral-latest`: currently points to `codestral-2501`. ## Model deprecation ### Overview @@ -82,4 +82,5 @@ To prepare for model retirements and version upgrades, we recommend that custome | Mistral Medium | | :heavy_check_mark: | Ideal for intermediate tasks that require moderate reasoning | 32k | `mistral-medium-2312`| 23.12 |2024/11/25|2024/11/30|2025/03/30|`mistral-small-latest`| | Mistral Small 24.02| | :heavy_check_mark: | Our latest enterprise-grade small model with the first version released Feb. 2024 | 32k | `mistral-small-2402` | 24.09| 2024/11/25|2024/11/30|2025/03/30| `mistral-small-latest`| | Mistral Large 24.02 | | :heavy_check_mark: |Our top-tier reasoning model for high-complexity tasks with the the first version released Feb. 2024. Learn more on our [blog post](https://mistral.ai/news/mistral-large/) | 32k | `mistral-large-2402`| 24.02|2024/11/25|2024/11/30|2025/03/30| `mistral-large-latest` | -| Mistral Large 24.07 | :heavy_check_mark:
[Mistral Research License](https://mistral.ai/licenses/MRL-0.1.md) | :heavy_check_mark: |Our top-tier reasoning model for high-complexity tasks with the the second version released July 2024. Learn more on our [blog post](https://mistral.ai/news/mistral-large-2407/) | 128k | `mistral-large-2407`| 24.02|2024/11/25|2024/11/30|2025/03/30| `mistral-large-latest`| +| Mistral Large 24.07 | :heavy_check_mark:
[Mistral Research License](https://mistral.ai/licenses/MRL-0.1.md) | :heavy_check_mark: |Our top-tier reasoning model for high-complexity tasks with the the second version released July 2024. Learn more on our [blog post](https://mistral.ai/news/mistral-large-2407/) | 131k | `mistral-large-2407`| 24.02|2024/11/25|2024/11/30|2025/03/30| `mistral-large-latest`| +| Codestral |:heavy_check_mark:
[Mistral Non-Production License](https://mistral.ai/licenses/MNPL-0.1.md) | :heavy_check_mark: | Our cutting-edge language model for coding with the first version released [May 2024](https://mistral.ai/news/codestral/) | 32k | `codestral-2405` | 24.05| 2024/12/02|2024/12/02|2025/03/30| `codestral-latest`| diff --git a/docs/guides/tokenization.mdx b/docs/guides/tokenization.mdx index 46d48b4..5c0a172 100644 --- a/docs/guides/tokenization.mdx +++ b/docs/guides/tokenization.mdx @@ -58,7 +58,7 @@ Tekken proved more proficient in compressing text for approximately 85% of all l Our tokenization vocabulary is released in the https://github.com/mistralai/mistral-common/tree/main/tests/data folder. Let’s take a look at the vocabulary of our v3 tekken tokenizer. #### Vocabulary size -Our vocabulary consists of 130k vocab + 1k control tokens. We can use up to 131k tokens and we current use 128k tokens. +Our vocabulary consists of 130k vocab + 1k control tokens. #### Control tokens Our vocabulary starts with 14 control tokens, which are special tokens we use in the encoding process to represent specific instructions or indicators: