Mistral Medium (latest)

Mistral Medium 3.5 is Mistral AI's first flagship "merged" model, a dense 128B parameter design that unifies instruction-following, reasoning, and coding within a single set of weights. Released as open weights under a modified MIT license, it is positioned as the new default model powering Mistral Vibe and Le Chat, replacing Devstral 2 in Mistral's coding agent. The architecture emphasizes long-horizon task execution, with the ability to call multiple tools reliably in parallel and produce structured output that downstream code can consume directly.

On practical benchmarks, Mistral Medium 3.5 scores 77.6% on SWE-Bench Verified, placing it ahead of Devstral 2 and larger competitors like Qwen3.5 397B A17B, while also achieving a strong 91.4 on τ³-Telecom for agentic telecom tasks. The model was engineered for real-world deployment efficiency, running self-hosted on as few as four GPUs, and its reasoning effort is configurable per request so the same weights can serve both quick chat replies and complex agentic runs. This combination of a large context window, strong coding and agentic benchmarks, and self-hostable scale makes it well-suited for teams building asynchronous cloud coding agents and multi-step productivity workflows that require sustained tool use.

Tempr Gatewaymistral/mistral-medium-latestmistral-medium

Open provider page Provider docs

Quick Info

Provider: Tempr Gateway
Model key: mistral/mistral-medium-latest
Release date: Apr 29, 2026
Last updated: Apr 29, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $1.50
Output token cost: $7.50

Limits

Output tokens: 262,144 tokens
Context window: 262,144 tokens

Transparent token rates

Compare Mistral Medium (latest) pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Mistral Medium (latest)

Eden AI

Official sourceAnnouncement

Mistral AI's May 22, 2026 blog post announces Mistral Medium 3.5 as a 128B dense open-weight model with a 256k context window that merges instruction-following, reasoning, and coding in a single set of weights, released under a modified MIT license. The post reports pricing of $1.50 per million input tokens and $7.50 p The announcement describes benchmark results including 77.6% on SWE-Bench Verified and 91.4 on τ³-Telecom for agentic telecom tasks, positioning Mistral Medium 3.5 ahead of Devstral 2 and larger competitors like Qwen3.5 397B A17B. For developers consuming mistral-medium-latest via Eden AI or any other provider, this po

Eden AI

Official sourceRelease Notes

Mistral's official documentation changelog records that on April 27, 2026, the company released Mistral Medium 3.5 under the model identifier mistral-medium-3-5, a release that falls squarely within the mistral-medium family that the Eden AI listing's "mistral/mistral-medium-latest" alias resolves to. This first-party The same changelog excerpt places the Mistral Medium 3.5 release in the context of broader Mistral activity, including the May 27, 2026 launch of Vibe as the company's unified agent (covering productivity, code, and chat modes) and earlier OCR and Leanstral model updates. While these surrounding items are not medium-fa