Ministral 3: December 2025 Small Open Models in Three Sizes. This reference distinguishes the original release from hosted endpoints, downloadable weights and subsequent version changes. It is written for people deciding whether an exact model or checkpoint is suitable for a real application, not for leaderboard marketing.
What the sources actually establish
Mistral’s December 2, 2025 Mistral 3 announcement included three smaller dense Ministral 3 variants at 3B, 8B and 14B parameters, all described under Apache 2.0. The official changelog records distinct API IDs beginning `ministral-3b-2512`, `ministral-8b-2512` and `ministral-14b-2512`. Calling the family simply `ministral-3` is useful for editorial organization but is not a valid claim of one immutable endpoint.
A useful application and evaluation scenario
A field-service mobile assistant may prefer a small model for drafting forms or classifying short maintenance notes. Evaluate on the actual mobile, laptop or enterprise hardware available, including device wake-up latency, multilingual data and offline constraints. A 3B checkpoint can be attractive for short tasks, but hardware fit and correctness cannot be inferred from parameter count without quantization and context-size measurements.
Limits, risks and uncertainties
The lineup belongs to a specific 2025 release. Earlier Ministral 3B/8B products announced in 2024 are not the same models despite overlapping numbers. Licenses should be tied to the exact original artifact and any later hosted packaging. A model that is fast for short text may still hallucinate under document questions or fail strict JSON requirements.
Alternatives and the decision rule
Begin with the smallest checkpoint that passes a reviewed acceptance suite; increase size only when the improvement warrants additional operational complexity. Store the three exact dated API IDs independently in a registry, and keep the family article separate from Mistral Large 3.
A reproducible test before adoption
Prepare a small held-out set from data the organization is authorized to process. Include simple, ambiguous and deliberately unanswerable requests; specify the expected answer or evidence and a human-reviewed scoring rule. Record model identifier, endpoint and region, tokenizer or SDK version, prompt, tool settings, output budget and whether external retrieval was allowed. Do not present a provider-supplied benchmark as an independently measured result. Separate model mistakes from retrieval, OCR and application validation failures.
Operational checklist and dated evidence
A production transition must test 429 responses, timeouts, refusals, invalid structured output and model-unavailable errors. Keep a reversible rollout or safe fallback instead of silently substituting a model behind an old identifier. Measure latency and usage on the actual workload before planning budgets; join dated token quantities to the AI 10 price ledger. Verify enterprise residency, region, support entitlement and data retention with the actual contract, not a public release story.
Sources & Further Reading
- mistral.ai — documentation/reference (supplier-provided reference; not independently checked in this pass)
- Mistral changelog (retrieved from first-party public documentation on 2026-10-25; this is not a deployment test)
Source documentation establishes vendor-published facts, not independently measured model quality, tenant access, contractual rights or current prices. Use a dated pricing ledger and an account-specific entitlement check for deployment decisions.