#llm · Cascadity

MVMaker Views #ai-coding #cybersecurity #europe #Hacker-News #llm #mistral #model-release #open-weights

Mistral releases the weights of Mistral Large 4, its 1-trillion-parameter flagship

Mistral's post has the benchmark charts, the reasoning modes and the license terms; the thread argues over what 'open' means here.

Mistral released the weights of Mistral Large 4 (nicknamed "Le Chonk"), a 1-trillion-parameter multimodal model that entered public preview earlier in October. The company says three more weeks of reinforcement learning since the preview improved coding and defensive cybersecurity: it reports leading open-weight scores on Cognition's FrontierCode 2.0 (19.7%) and Terminal-Bench 4.0 (41%), and 51% on the Artificial Analysis Cyber Index, with a 96.6% refusal rate on malicious cyber prompts. These are vendor-reported results. It adds low, medium and high reasoning modes and says the model was trained from scratch on European infrastructure.

Weights are on Hugging Face under the "Mistral Ecosystem Open-Weight license," a modified MIT license: free to use, modify and commercialize, except that model-as-a-service providers need a separate commercial license. API pricing is $0.70 per million input tokens and $2.10 per million output tokens.

On Hacker News, a commenter noted the price is roughly half the preview's, another argued the license makes it "weights available" rather than open source, and several said its main draw is European provenance.

Read it at the source

Mistral AIMistral Large 4 (Le Chonk) is now open weight | Mistral ↗ Hacker NewsDiscussion: We're releasing the weights of Mistral Large 4 Independent ↗

Cascadic Analysis 2

UndertowBenchmark scores can overstate real-world reliability

What hidden risk could pull against this, even if the news is good?

Most of the headline numbers come from Mistral's own release page, and the 96.6% refusal rate is an average over public test sets of malicious prompts (JailbreakBench, StrongREJECT, AgentHarm). Known prompts are what a determined user is least likely to send. Mistral itself says it sees "potential for uplift" toward real attacks and is adjusting its guardrails.

UndertowAutonomous cyber capability scales attackers enormously

What hidden risk could pull against this, even if the news is good?

Mistral keeps its specialized cyber variant behind vetting and contracts, but the general model, at 51% on a vulnerability finding-and-patching index, ships as open weights that the license lets anyone modify and fine-tune. Its release page says built-in guardrails separate defensive from offensive tasks, without saying how they hold up once the weights are self-hosted. A downloaded model can't be recalled if that line proves thin.

Rabbit Holes 2

Open-source artificial intelligence

After exploring Mistral Large 4 (Le Chonk) is now open weight | Mistral, for the debate over what counts as 'open', read Open-source artificial intelligence.

Covers the Open Source Initiative's definition of open-source AI (free to use, study, modify and share, including data, code and parameters) and how it differs from releasing weights alone.

Wikipedia · Wade · 5 min

0
Open weights

After exploring Mistral Large 4 (Le Chonk) is now open weight | Mistral, for what 'open weights' does and doesn't mean, read Open weights.

Explains open weights: publicly released learned parameters of an AI model, which let others download and run it, with modification and redistribution depending on the license.

Wikipedia · Wade · 5 min

0

MVMaker Views #benchmarks #deepseek #Hacker-News #llm #open-weights

DeepSeek releases V4.1-Flash-1017, an MIT-licensed "minor update" with sizable benchmark gains

The model card has the full benchmark table against GLM-5.3, GPT-5.6 Sol and Opus 5.

DeepSeek published DeepSeek-V4.1-Flash-1017 on Hugging Face, calling it a minor update to V4.1-Flash that brings substantial improvements in reasoning, coding, agentic tasks and visual benchmarks while cutting task time and output length. Its table shows, for example, Terminal-Bench 4.0 rising from 31.2 to 42.3 and DeepSWE v1.1 from 74.2 to 77.0, ahead of GPT-5.6 Sol on several tests and behind Opus 5 on others. The model has a 1M-token context window, weights are MIT-licensed, and DeepSeek released deepseek-recipe, Rust libraries with Python bindings for its prompt format.

On Hacker News, users said the model feels better than its benchmarks because of speed, visible reasoning and fewer refusals; others posted informal head-to-head tests and cautioned that single runs prove little. Some joked about calling big gains a minor update; a side debate asked whether US labs want regulation that would hurt open models.

Read it at the source

Hugging Face (DeepSeek)deepseek-ai/DeepSeek-V4.1-Flash-1017 · Hugging Face Independent ↗ DeepSeekDeepSeek-V4.1-Flash-1017: Wish-Driven Programming Made Real | DeepSeek API Docs Independent ↗ Hacker NewsDiscussion: DeepSeek-v4.1-Flash-1017 Independent ↗

Cascadic Analysis 2

UndertowBenchmark scores can overstate real-world reliability

What hidden risk could pull against this, even if the news is good?

The gains, such as Terminal-Bench 4.0 from 31.2 to 42.3, come from DeepSeek's own table on its model card. Hacker News users posted head-to-head tests of their own, and others cautioned that single runs prove little; how the model holds up on a team's own tasks still has to be measured there.

UndertowLoss of meaningful human control as capability and autonomy increase

What hidden risk could pull against this, even if the news is good?

The weights are MIT-licensed and downloadable, and some Hacker News users liked the model partly for its fewer refusals. Once the weights are out, how the model is fine-tuned, what it refuses and where it runs is up to whoever downloaded it; DeepSeek can't recall or patch copies already in circulation.

Rabbit Holes 1

Mixture of experts

After exploring deepseek-ai/DeepSeek-V4.1-Flash-1017 · Hugging Face, for the architecture idea behind models like this, read about Mixture of experts.

A machine-learning technique that splits a problem among multiple expert networks, a form of ensemble learning.

Wikipedia · Swim · 30 min

0