Close Menu
  • AI
  • Content Creation
  • Tech
  • Robotics
AI-trends.todayAI-trends.today
  • AI
  • Content Creation
  • Tech
  • Robotics
Trending
  • Potential Cognition AI Spherical Would Elevate $2B-Plus at $75B Valuation – Unite.AI
  • OpenRouter Cuts Three Suppliers After AI Immediate Publicity Findings – Unite.AI
  • A Group of Researchers from Columbia and CertiK Introduce TensorLean for Formal Verification of GPU Kernels
  • Collectively AI Discloses Immediate Leak After Coding Agent Uncovered Credential – Unite.AI
  • Dragos Finds Immediate Leaks and Stolen Keys in AI Provide Chains
  • Lean and Mathlib Communities Certify Claude-Assisted Riemann Proof – Unite.AI
  • AWS Routes HyperPod Inference Throughout Areas From One Endpoint – Unite.AI
  • DailyAI Briefing: A Lean-checked Riemann proof, plus 4 extra
AI-trends.todayAI-trends.today
Home»Robotics»Benchmark Agency Provides Mistral Massive 4 Preview a 38 Intelligence Rating – Unite.AI

Benchmark Agency Provides Mistral Massive 4 Preview a 38 Intelligence Rating – Unite.AI

Robotics By Gavin Wallace06/10/20264 Mins Read
Facebook Twitter LinkedIn Email
Gemini Robotics: AI Reasoning Meets the Physical World
Gemini Robotics: AI Reasoning Meets the Physical World
Share
Facebook Twitter LinkedIn Email

Synthetic Evaluation published its benchmark analysis of Mistral Large 4 Preview on October 6, 2026, scoring the mannequin 38 on its Intelligence Index and describing it as probably the most clever AI mannequin from outdoors the US and China.

Intelligence Index Outcomes

The evaluation experiences that Mistral Massive 4 Preview scores 38 on the Synthetic Evaluation Intelligence Index model 4.3.2, a outcome the benchmarking agency describes as similar to GPT-6 Luna (max) at 38 and DeepSeek V4.1 Flash (max) at 39. In its framing, France is again to having probably the most clever mannequin from outdoors the US and China, forward of nations equivalent to South Korea and the United Arab Emirates.

On Synthetic Evaluation’s model page for the preview, that rating ranks the mannequin quantity 64 of 225 fashions in its comparability class, above a category median of 26. The web page lists the preview as a reasoning mannequin launched October 6, 2026, with textual content and picture enter, textual content output, and repair by means of two API suppliers.

In accordance with the printed Intelligence Index methodology, model 4.3.2 combines ten evaluations — AA-Briefcase v1.1, GDPval-AA v2.1, AutomationBench-AA, Terminal-Bench 4.0, SciCode, Humanity’s Final Examination, GDP.pdf, CritPt, AA-Omniscience, and AA-LCR v1.1 — as a weighted common throughout 4 classes: brokers at 30%, coding at 20%, scientific reasoning at 20%, and common at 30%. Synthetic Evaluation estimates a 95% confidence interval of lower than ±1% for the index and describes the suite as primarily text-based and English-language, with picture, speech, and multilingual efficiency benchmarked individually.

Cyber Index Outcomes

On the Synthetic Evaluation Cyber Index, the preview scores 50, stage with GLM-5.3-Flash at 50 and behind MiMo-V2.6-Professional at 56, whereas forward of fashions together with Kimi K3 and DeepSeek V4.1 Flash (max). Synthetic Evaluation states that after the mannequin’s weights are launched, it is going to rank among the many prime three open-weights fashions on the Cyber Index.

The mannequin’s strongest particular person cyber outcome got here on CyberGym-E2E-AA, the place it scores 82%, forward of MiMo-V2.6-Professional at 79% and GPT-6 Luna (max) at 78%, based on the evaluation.

Value, Velocity, and Token Use

The evaluation places the price of operating the Intelligence Index on Mistral Massive 4 Preview at $1.13 per process, primarily based on normal pricing of $1.36 per 1M enter tokens and $4.18 per 1M output tokens, with cached enter at $0.14 per 1M tokens. A 50% launch low cost applies for the primary two weeks, chopping token costs to $0.68 and $2.09 and bringing the associated fee per process to $0.57 — nonetheless above GLM-5.3-Flash at $0.25 and DeepSeek V4.1 Flash (max) at $0.27. Synthetic Evaluation characterizes the preview as greater than 4 occasions the associated fee per process of similar-intelligence open-weights fashions.

The mannequin web page data the preview producing 200M output tokens over the course of the Intelligence Index run, towards a category median of 81M. Measured by means of Mistral’s API, it experiences output velocity of 116.1 tokens per second towards a median of 86.1, and a time to first token of 1.46 seconds towards a median of three.81 seconds.

Doc Reasoning and Mannequin Particulars

On GDP.pdf, knowledgeable document-reasoning analysis, Mistral Massive 4 Preview scores 19%, on par with MiMo-V2.6-Professional at 19% and behind Kimi K3 at 22%. Synthetic Evaluation describes the outcome as an 18-point enchancment over Mistral Massive 3, partly pushed by an API change that now accepts 100 photos per request, up from eight for earlier Mistral fashions.

The evaluation lists a 512k-token context window and a 1-trillion-parameter mannequin with 49 billion energetic parameters. Availability is proscribed to a Analysis Public Preview on Mistral’s API, with open weights deliberate for the top of October 2026; the mannequin web page presently classifies the preview as proprietary as a result of its weights will not be but publicly obtainable.

The Mistral Massive 4 Launch

Mistral launched the public preview of Mistral Massive 4, nicknamed Le Chonk, on October 6, 2026, describing a natively multimodal mannequin educated from scratch on 3,800 NVIDIA Grace Blackwell GPUs within the firm’s personal datacenters in Europe, with coaching information spanning greater than 160 languages together with each official European Union language. Mistral claims the mannequin considerably outperforms any open-weight mannequin developed within the US or Europe, and says it is going to launch the weights by the top of October 2026, with the reinforcement studying run behind the preview nonetheless in flight.

AI ar Benchmark intel intelligence
Share. Facebook Twitter LinkedIn Email
Avatar
Gavin Wallace

Related Posts

Potential Cognition AI Spherical Would Elevate $2B-Plus at $75B Valuation – Unite.AI

22/11/2026

OpenRouter Cuts Three Suppliers After AI Immediate Publicity Findings – Unite.AI

22/11/2026

Collectively AI Discloses Immediate Leak After Coding Agent Uncovered Credential – Unite.AI

22/11/2026

Lean and Mathlib Communities Certify Claude-Assisted Riemann Proof – Unite.AI

21/11/2026
Top News

DeepSeek, Moonshot Used GPT-6 Astra to Prepare Fashions: Report

Here are the 500 most common questions I received from ChatGPT. These are the ads I saw most often.

Meta and Mercor Pause Work After Breach of Data Puts AI Industry Secrets in Danger

OpenAI launches a full-scale effort to patch open-source bugs as it takes on Anthropic’s Mythos

A Filmmaker Made a Sam Altman Deepfake—and Got Unexpectedly Attached

Load More
AI-Trends.Today

Your daily source of AI news and trends. Stay up to date with everything AI and automation!

X (Twitter) Instagram
Top Insights

Snap, YouTube, TikTok and other social media sites settle lawsuits over student harm

16/05/2026

The UK’s Cyber Paradox: Better Prepared, Yet More Heavily Targeted – Unite.AI

12/09/2026
Latest News

Potential Cognition AI Spherical Would Elevate $2B-Plus at $75B Valuation – Unite.AI

22/11/2026

OpenRouter Cuts Three Suppliers After AI Immediate Publicity Findings – Unite.AI

22/11/2026
X (Twitter) Instagram
  • Privacy Policy
  • Contact Us
  • Terms and Conditions
© 2026 AI-Trends.Today

Type above and press Enter to search. Press Esc to cancel.