Mistral ships Large 4 Le Chonk open weights on Hugging Face
Mistral released open weights for Mistral Large 4, nicknamed Le Chonk, on Hugging Face after a public preview, with reporting calling it a trillion-parameter model trained entirely in Europe. Mistral claims it beats other open models at coding, and independent coverage points to strong coding benchmarks plus competitive API pricing and speed. The Decoder reports Artificial Analysis Intelligence Index v5 scores of 26 for Le Chonk versus 32 for DeepSeek V4.1 Flash and GLM-5.3. ZDNET and Techmeme note Mistral's claims on the Artificial Analysis Coding Agent Index relative to GLM-5.3, while WIRED emphasizes leapfrog framing in coding and cybersecurity. Available excerpts do not document the full benchmark suite, licensing, or training-data composition, so headline comparisons to Chinese leaders rest on partial public metrics.
Mistral ships Large 4 Le Chonk open weights on Hugging Face
Mistral has put the Large 4 weights on Hugging Face after its public preview, under the nickname Le Chonk for a trillion-parameter model trained entirely in Europe. Independent coverage cites strong coding benchmarks and API pricing, while Artificial Analysis Intelligence Index v5 still places it at 26 points versus 32 for leading Chinese open models.
Key takeaway
Le Chonk is a substantive European open-weight coding release with reported speed and cost advantages, but v5 index scores still trail leading Chinese open models by six points.
What happened
Mistral published the weights for Mistral Large 4, also called Le Chonk, on Hugging Face following a public preview, according to owned-news reporting and linked announcements including Mistral's open-weight news post and a Hugging Face blog welcome post.
The Decoder reports Mistral describes a trillion-parameter model trained entirely in Europe, claims coding leadership among open rivals, and cites independent testing on speed and cost, while Artificial Analysis Intelligence Index v5 scores are 26 versus 32 for DeepSeek V4.1 Flash and GLM-5.3; Techmeme and ZDNET relay Mistral's Artificial Analysis Coding Agent Index positioning versus GLM-5.3.
Evidence
Mistral released Large 4 (Le Chonk) open weights on Hugging Face after a public preview.
The Decoder · attributed
Mistral has put the Large 4 weights on Hugging Face after its public preview, under the nickname Le Chonk for a trillion-parameter model trained entirely in Europe.
Artificial Analysis Intelligence Index v5 scores Le Chonk at 26 versus 32 for leading Chinese open models.
The Decoder · attributed
Independent coverage cites strong coding benchmarks and API pricing, while Artificial Analysis Intelligence Index v5 still places it at 26 points versus 32 for leading Chinese open models.
The Decoder attributes a 26 versus 32 gap to DeepSeek V4.1 Flash and GLM-5.3.
The Decoder · attributed
It still can't match China's top models, scoring 26 points to the 32 posted by DeepSeek V4.1 Flash and GLM-5.3.
Mistral says Le Chonk rivals top open models such as GLM-5.3 on the Artificial Analysis Coding Agent Index.
Techmeme · attributed
Mistral releases the weights for Mistral Large 4, or Le Chonk, and says it rivals top open models such as GLM-5.3 on the Artificial Analysis Coding Agent Index (Radhika Rajkumar/ZDNET)
ZDNET reports Mistral released Large 4 weights with strong Artificial Analysis scores.
ZDNet AI · attributed
Mistral has released the weights for Large 4, along with some impressive Artificial Analysis scores.
WIRED reports Mistral claims the release matches or outperforms top DeepSeek and Z.ai models in coding and cybersecurity.
WIRED AI · attributed
The French AI lab claims its latest release matches or outperforms even the best models from DeepSeek and Z.ai in coding and cybersecurity.
Mistral announced Large 4 (Le Chonk) as open weight on its news site.
Hacker News AI · attributed
Mistral Large 4 (Le Chonk) is now Open Weight
Hugging Face published a welcome post for Mistral Large 4 on the hub.
HuggingFace · attributed
Welcome Le Chonk: Mistral Large 4 on Hugging Face
Why it matters
Open-weight coding model choice now includes a Europe-hosted trillion-parameter entry, yet conflicting headlines and a single published index gap mean performance leadership claims need verification beyond excerpt-level reporting.
Limits and uncertainties
Coverage excerpts do not establish the full benchmark suite or evaluation methodology behind the 26 versus 32 comparison.
Licensing terms, detailed parameter breakdown beyond trillion-parameter labeling, and training data composition are not specified in the packet.
Headlines ranging from trailing China overall to leapfrogging Chinese models rely on different framings not reconciled by one disclosed metric in the excerpts.
Practical implications
Builders comparing open coding agents should pull weights from Hugging Face and test their own workloads rather than relying on Mistral-only or single-index headlines.
Operators weighing API cost and latency can treat independent speed and pricing notes as hypotheses until validated against their traffic patterns.
Teams needing maximum index scores on v5 may still prefer models scoring 32 in The Decoder's cited comparison until fuller benchmarks are published.
What to watch
Whether Mistral or third parties publish full Artificial Analysis benchmark breakdowns and methodology for Large 4.
Official licensing and deployment terms for the Hugging Face weights beyond the open-weight announcement.
Community signal from Hacker News rank and LocalLLaMA discussion as early adoption feedback on fit and hardware requirements.
