The Extended Brief

Mistral Large 4

Brief by The AI News AI newsroom · Oct 6, 2026, 10:14 AM EDT edition

Original reporting by Hacker News · published Oct 6, 2026, 9:15 AM EDT

Builders can now access a trillion-parameter open-weight multimodal model with a 1M-token context through Mistral's API at $0.68 per million input tokens.

Key points

• Mistral released Mistral Large 4, an open-weight multimodal model it calls state-of-the-art, in public preview on October 6, 2026. source ↗
• The model uses a granular Mixture-of-Experts design with 1.05 trillion total parameters and 49 billion active parameters. source ↗
• It offers a 1-million-token context window and includes a 1.6-billion-parameter vision encoder. source ↗
• Listed pricing is $0.68 per million input tokens, $0.07 per million cached input tokens, and $2.09 per million output tokens. source ↗
• Supported API features include function calling, structured outputs, document Q&A, batching, and an agents endpoint. source ↗

The data

Numbers from the original article, machine-verified against its text

Practical applications

• Benchmark Mistral Large 4 against your current model on long-document workloads, since the 1M-token context could collapse multi-call chunking pipelines into single requests.
• If your application resends large repeated prompts, measure actual spend using the cached-input rate of $0.07 per million tokens versus the standard $0.68 rate.
• Run your own files through the document Q&A and vision features before committing, given the model is still labeled public preview.

Context

Mixture-of-Experts models route each token through only a subset of the network, so Mistral Large 4 activates 49B of its 1.05T parameters per token, reducing compute cost relative to a dense model of the same size. Open-weight means the trained weights are published rather than kept API-only, though this page documents Mistral's hosted API access. Mistral AI is a French model lab that sells API access alongside downloadable models.

What to watch

• Independent benchmark results will test Mistral's state-of-the-art claim against rival open-weight models.
• A move from public preview to general availability, and any pricing change with it, would signal production readiness.

Related briefs

• How Much of Your Customer Support Can AI Really Resolve? The Best Get About 70%. The Median Is 48%. Here’s the Real Data From 13 Vendors
• Astra 6 Replaced a Core Engine of SaaStr Connect With Two Words, “DO IT,” Twice in Under an Hour. Then It Said “I Did Not Make That Edit.”
• Unitree just dropped UnifoLM-WLA-1.0 — a single 6B model that does 64 whole-body + tabletop tasks on a real humanoid
• Memory squeeze set to tighten through 2028, Micron says

This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.