← The Wire

Models2026-06-25 · source-backed

Mistral Medium 3.5 hits 77.6% on SWE-Bench Verified and becomes the Le Chat default.

Confidence
source-backed
Sources
1
Redaction
passed

Story

On June 24 Mistral set Medium 3.5 as the default in Le Chat, reporting 77.6% on SWE-Bench Verified, ahead of Devstral 2 (Mistral via Releasebot). A European, open-leaning option clearing roughly 78% on SWE-Bench Verified is a credible alternative to US frontier coding models, especially stacked with the scoped-key Connectors. The agentic-coding field is getting genuinely crowded at the "good enough" tier, which is great for builders and rough for anyone trying to charge frontier prices.

Related stories

Each link below shares sources, entities, or timing with this story.

  1. Shared entities / Same source / Shared topic

    Mistral pairs scoped-key Connectors with a 10 MW EU inference facility.

    Both cover Connectors, European, Mistral, Releasebot; cite the same source (Mistral via Releasebot); overlapping topics (connector, mistral).

  2. Shared entities / Shared topic / Earlier coverage

    Devstral 2 is open source, 123B params, hits 72.2% SWE-bench, and costs a seventh of Sonnet

    Both cover Devstral, Le Chat, Mistral, SWE; overlapping topics (coding, frontier, mistral, swe-bench); earlier Devstral coverage from 2026-06-02.

  3. Mistral 3 + Devstral 2 + Vibe CLI 2.0

    Both cover Devstral, Mistral, SWE; overlapping topics (coding, mistral); earlier Devstral coverage from 2026-03-01.

  4. Shared entities / Shared topic / Earlier coverage / Tension

    Kimi K2.6: A 1-Trillion-Parameter Open-Weights Model Just Beat GPT-5.4 on SWE-Bench Pro

    Both cover Bench Verified, SWE; overlapping topics (coding, frontier, swe-bench); earlier Bench Verified coverage from 2026-04-21.

  5. Shared entities / Same source domain / Shared topic / Earlier coverage

    Vibe Coding & AI Development

    Both cover Medium, Releasebot; reported by the same outlet (releasebot.io); overlapping topics (coding, medium).

  6. Shared entities / Shared topic / Earlier coverage / Tension

    Cohere open-sourced a 30B coding model that runs on a single H100. Agentic coding without the frontier bill.

    Both cover Bench Verified, SWE; overlapping topics (coding, frontier); earlier Bench Verified coverage from 2026-06-10.

  7. Shared entities / Shared topic / Earlier coverage

    Mistral Medium 3.5: 128B dense open-weight model with 256K context under modified MIT.

    Both cover Mistral, Mistral Medium; overlapping topics (anyone, coding, frontier); earlier Mistral coverage from 2026-04-30.

  8. Mistral 3 Family + Devstral 2 + Vibe CLI — Apache 2.0 Open-Source Coding Agent

    Both cover Devstral, Mistral; overlapping topics (builder, coding, mistral); earlier Devstral coverage from 2026-02-28.

Source trail

Entities

Provenance

AI generated
no
Story unit
2026-06-25-mistral-medium-3-5-hits-77-6-on-swe-bench-verified-and-becomes-the-le-chat-default
Labels
source-backed, canonical briefing excerpt