Mistral Small 3
Mistral Small 3 is a model released by Mistral AI under the Apache 2.0 licence, with both a pretrained and an instruction-tuned checkpoint published as open weights. It is distributed through a serverless API on Mistral's la Plateforme, through the vendor's on-prem and VPC deployments, and as weights that can be downloaded and run locally. At launch it was made available on Hugging Face, Ollama, Kaggle, Together AI, Fireworks AI and IBM Watson X; NVIDIA NIM, Amazon SageMaker, Groq, Databricks and Snowflake are stated as coming soon rather than live. The vendor describes low-latency function calling for agentic workflows and fine-tuning to build domain specialists, naming legal advice and medical diagnostics among the intended fields, and reports customers evaluating the model in financial services, healthcare, robotics, automotive and manufacturing. Its only published evaluation is a side-by-side run over more than 1,000 proprietary coding and generalist prompts that Mistral commissioned from an external vendor: a vendor-commissioned quality comparison, not an independent audit, and not a safety or dangerous-capability evaluation. Nothing on the announcement page checked describes runtime limits on an agent's scope, spend or authority, an audit trail, a compliance certification, outcome-based settlement, or insurance or indemnity cover.
14 fields evidenced · 41 with no public information. Every value below links to the document it came from and the date we checked it.
Also appears in
Trust gap
How this is scoredAre scope, spend, and authority enforced while the agent runs?
No public evidence found. The Mistral Small 3 announcement page checked (retrieved 2026-08-04) describes function calling and deployment options but states no enforced limit on an agent's scope, spend or authority at run time.
How this gap gets closed →Is there a tamper-evident record of what it actually did?
No public evidence found. The announcement page checked describes no record of model or agent actions, reviewable or otherwise.
How this gap gets closed →Are compliance obligations attached per engagement?
No public evidence found. The announcement page checked names no certification, audit report or per-engagement compliance schedule. The Apache 2.0 licence is a distribution term, not a compliance control.
How this gap gets closed →Does payment depend on a verified result?
No public evidence found. The model is distributed as downloadable weights and a serverless API; the page checked describes no payment released against a verified outcome.
How this gap gets closed →Can the deployment be insured, and is the customer indemnified?
No public evidence found. The announcement page checked names no carrier, no cover, no published limits and no customer indemnity.
How this gap gets closed →Assurance
- Insurance available
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Insurance carriers
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Coverage limits
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Indemnification
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Liability cap
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Audit trail
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Tamper-evident log
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Explainability
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Runtime governance
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Permission scopes
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Compliance certifications
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Regulatory alignment
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Outcome-based pricing
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Settlement mechanism
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Dispute process
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- SLA terms
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Evaluation coverage
- side by side evaluations with an external third-party vendor on over 1k proprietary coding and generalist prompts
“We conducted side by side evaluations with an external third-party vendor, on a set of over 1k proprietary coding and generalist prompts.”
mistral.ai · checked Aug 4, 2026
Agency
- Autonomy level
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Human oversight
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Goal complexity
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Action space
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Operating environment
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Initiative
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
Safety
- Safety evaluations
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Red teaming
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Safety policy
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Usage restrictions
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Model or system card
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Incident reporting
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Third-party evaluations
- Vendor-commissioned: Mistral engaged an external third-party vendor to run side-by-side evaluations. Not an independent audit published by that vendor.
“We conducted side by side evaluations with an external third-party vendor”
mistral.ai · checked Aug 4, 2026 - Data handling
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
Practicality
- Pricing model
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Price point
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Availability
- Publicly announced and available as of the 2026 launch post; weights downloadable and served via API.
“Today we’re introducing Mistral Small 3”
mistral.ai · checked Aug 4, 2026 - Deployment options
- serverless API, on-prem, VPC, local deployment
“model weights will be available to download and deploy locally, and free to modify and use in any capacity. These models will also be made available through a serverless API on la Plateforme , through our on-prem and VPC deployments”
mistral.ai · checked Aug 4, 2026 - Integrations
- Available at launch on Hugging Face, Ollama, Kaggle, Together AI, Fireworks AI and IBM Watson X; NVIDIA NIM, Amazon SageMaker, Groq, Databricks and Snowflake are stated as coming soon, not live.
“We are also excited to collaborate with Hugging Face, Ollama, Kaggle, Together AI, and Fireworks AI to make the model available on their platforms starting today: Hugging Face ( base model ) Ollama Kaggle Together AI Fireworks AI IBM Watson X Coming soon on NVIDIA NIM, Amazon SageMaker, Groq, Databricks and Snowflake”
mistral.ai · checked Aug 4, 2026 - Supported regions
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Support model
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
Foundation models
- Base models
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Model provider
- Mistral AI
“Mistral Small 3 | Mistral AI”
mistral.ai · checked Aug 4, 2026 - Model swappable
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Open weights
- Apache 2.0 license
“We’re releasing both a pretrained and instruction-tuned checkpoint under Apache 2.0.”
mistral.ai · checked Aug 4, 2026 - Fine-tuning
- fine-tuned to specialize in specific domains
“Fine-tuning to create subject matter experts: Mistral Small 3 can be fine-tuned to specialize in specific domains”
mistral.ai · checked Aug 4, 2026 - Context window
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
Ecosystem
- Protocols supported
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Tool use
- handle rapid function execution
“Low-latency function calling: Mistral Small 3 is able to handle rapid function execution when used as part of automated or agentic workflows.”
mistral.ai · checked Aug 4, 2026 - Multi-agent
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- API access
- serverless API on la Plateforme
“These models will also be made available through a serverless API on la Plateforme”
mistral.ai · checked Aug 4, 2026 - Open source
- Apache 2.0 license
“released under the Apache 2.0 license”
mistral.ai · checked Aug 4, 2026 - Marketplace presence
- Hugging Face, Ollama, Kaggle, Together AI, Fireworks AI
“We are also excited to collaborate with Hugging Face, Ollama, Kaggle, Together AI, and Fireworks AI to make the model available on their platforms starting today”
mistral.ai · checked Aug 4, 2026
Impact
- User base
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Deployment scale
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
- Target sectors
- Financial services, Healthcare, Robotics, automotive, manufacturing
“Our customers are evaluating Mistral Small 3 across multiple industries, including: Financial services customers for fraud detection Healthcare providers for customer triaging Robotics, automotive, and manufacturing companies for on-device command and control”
mistral.ai · checked Aug 4, 2026 - High-risk domains
- legal advice, medical diagnostics
“Fine-tuning to create subject matter experts: Mistral Small 3 can be fine-tuned to specialize in specific domains, creating highly accurate subject matter experts. This is particularly useful in fields like legal advice, medical diagnostics, and technical support, where domain-specific knowledge is essential.”
mistral.ai · checked Aug 4, 2026 - Documented incidents
- no public informationmistral.ai · checked Aug 4, 2026 · source did not state this
Compared with
Side-by-side on the fields where Mistral Small 3 and the other entry are both on record and actually differ.