Mistral Large 4 Review: France Back in Third Place, but Value Questioned
On Oct 6, Artificial Analysis published its full review of Mistral Large 4 Research Public Preview: a 1T-parameter model (49B active) scoring 38 on the Intelligence Index, placing it near the Gemini/GPT tier, which the evaluator framed as France rejoining the strongest models outside the US and China. Weights are planned to open-source by month's end, but high pricing and token-efficiency concerns have sparked value-for-money controversy.
Confirmed
- Specs and overall: 1T total parameters (49B active), Intelligence Index 38, near the Gemini/GPT tier; full results page is live
- Country ranking: at 38, France rises to third, behind the US (Claude Opus 5.5, 58) and China (MiMo-V2.6-Pro, 46)
- Pricing: $1.36/$4.18 per million input/output tokens, cached input $0.14; 50% off for the first two weeks brings per-task cost to $0.57, but the full-rate per-task cost is about $1.13, roughly 4x comparable open models
- Value comparison: $1.13 per task for 38 points vs MiMo v2.6 Pro at $0.13 for 46; same score as GPT 6 Luna but 16x the cost ($0.07)
- Cyber capability: Cyber Index 50, tied with GLM-5.3-Flash and below MiMo-V2.6-Pro (56); evaluator expects a top-three open-model cyber ranking once weights ship
- Vulnerability test: against its 'strongest open model for cybersecurity' claim, a third-party VulnPR-100 run (100 PRs with known vulnerabilities, 1 hour each) found only 35/100, worse than Chinese open models; reshared by Mistral co-founder Guillaume Lample
- Researcher Yuchen Jin noted it clearly trails GLM-5.3 on the AA Intelligence Index, joking about an 'eval crisis'
- Independent evaluator haider found its output token usage per task is over 2x GPT-6 sol and Astra
- Community reaction: BLUECOW009 mocked it as 'le chaton skinny,' questioning the value of the 1T scale
Unconfirmed
- The month-end release of 1T-scale weights remains a stated plan only
- Third-party opinions on the 38 score diverge: LuminaBench called it poor for a 1T model, while others note parity with GLM 5.x-tier rivals
Why it matters
- An on-schedule open-sourcing of 1T-scale weights could reshape the open-model landscape and cement France as the third pole beyond the US and China
- The capability-cost mismatch is the core dispute: mediocre scores at multiples of rival cost and token usage may limit real adoption
- Cyber and other specialized capabilities still trail top Chinese open models like MiMo-V2.6-Pro, worth tracking after weights release
2026-10-06 ~ 2026-10-07 · 11 related posts
- Episode 1: Mistral to Launch New Flagship Model, Claims Edge Over Chinese Models(2026-10-06, 2 posts)
- Episode 2: Mistral Launches Mistral Large 4, a 1T-Parameter Open-Weight Flagship(2026-10-06, 86 posts)
- Episode 3: Mistral Large 4 Still in RL Training, Expected by Month's End(2026-10-06, 3 posts)
- Episode 4: Mistral Large 4 Review: France Back in Third Place, but Value Questioned(2026-10-06, 11 posts)
- Episode 5: Mistral Large 4 Teased on Hugging Face: 1T-Parameter Open-Weight Model Coming This Month(2026-10-06, 2 posts)
- Episode 6: Mistral Claims Large 4 Among Strongest Cybersecurity AI Models(2026-10-07, 2 posts)
- Episode 7: Mistral launches trillion-parameter multimodal Large 4, open weights coming this month(2026-10-10, 4 posts)
Primary sources
- Mistral Large 4 scores 38 on AA Intelligence Index, 1T open weights due end of October — ArtificialAnlys ·
- Mistral Large 4 pricing: $1.13 per task, 4x costier than similar open weights models — ArtificialAnlys ·
- France ranks third by country for frontier model intelligence on AA index — ArtificialAnlys ·
- [source] Mistral Large 4 scores 38 on AA Intelligence Index, 1T open weights due end of October — ArtificialAnlys · 2026-10-06
- [source] France ranks third by country for frontier model intelligence on AA index — ArtificialAnlys · 2026-10-06
- [source] Mistral Large 4 pricing: $1.13 per task, 4x costier than similar open weights models — ArtificialAnlys · 2026-10-06
- Mistral Large 4 scores 50 on Cyber Index, set for top-three open weights spot — ArtificialAnlys · 2026-10-06
- Full benchmark results for Mistral Large 4 Preview published by Artificial Analysis — ArtificialAnlys · 2026-10-06
- Mistral Large 4 trails GLM-5.3 on Artificial Analysis index, fueling 'eval crisis' talk — Yuchenj_UW · 2026-10-06
- Mistral Large 4 burns over 2x the output tokens per task vs GPT-6 sol — haider1 · 2026-10-06
- Mistral fights back: 38 points at $1.13/task, best open model in the West, top cyber score — rickasaurus · 2026-10-06
- Mistral Large 4 scores just 38 on AA intelligence index, panned for a 1T model — realmrfakename · 2026-10-06
- Mistral Large 4 scores just 38 on AA Intelligence Index despite 1T total params — BLUECOW009 · 2026-10-07
- Mistral Large 4 finds 35/100 vulnerable PRs for $53.91, trails Xiaomi and DeepSeek open models — GuillaumeLample · 2026-10-07