Mistral AI
Mistral Large 4
Mistral Large 4 entered public preview on October 6, 2026 — a 1.05T-parameter, natively multimodal MoE at $0.68/$2.09 per million tokens, with open weights promised for late October. Zero independent scores on this site's tracked benchmarks so far; Mistral's launch-chart claims are recorded on this page with their sourcing.
Mistral Large 4 benchmarks and pricing, every number sourced: 1 tracked Mistral Large 4 benchmark score (0 independently run, 1 still resting on a vendor’s own claim), priced at $0.68 per million input tokens and $2.09 per million output.
Mistral Large 4 architecture: MoE, natively multimodal (text + image input); 1.05T total parameters (49B activated per token); 1M-token context window.
- Released
- 2026-10-06
- Availability
- Public preview via API since 2026-10-06; open weights promised for late October 2026 while the model finishes partner red-teaming
- License
- proprietary
- Context window
- 1M tokensvendor-stated (docs.mistral.ai model page: "Context 1M"); Artificial Analysis's comparison summary separately lists 524k — the vendor spec is recorded here, AA's figure noted on the page
- Knowledge cutoff
- Not disclosed
- Verified
- source checked 2026-10-06
- Parameters
- 1.05T (49B active)
- Architecture
- MoE, natively multimodal (text + image input)
Mistral Large 4’s verified record
Mistral Large 4’s most-compared rival is Claude Sonnet 5.5: 1 not callable across their 1 shared comparison. Mistral Large 4 is announced at $0.68/$2.09 per 1M tokens (in/out) vs Claude Sonnet 5.5’s $2.00/$10.00.
Against the 29 head-to-head comparisons Mistral Large 4 shares with other tracked models: 0 real gaps, 0 inside the noise band, and 29 we will not call.
A gap counts for Mistral Large 4 only where both sides were run independently and the benchmark still separates models. Losses are listed alongside wins on purpose — Mistral Large 4 trails on 0 of them.
No verdict for Mistral Large 4 anywhere on DeepSWE ( nothing independently confirmed on both sides).
- None of Mistral Large 4’s coding comparisons are independently confirmed on both sides yet.
Mistral Large 4 announced pricing
$0.68 in / $2.09 out per 1M tokens — official pricing source
What Mistral Large 4 costs per job
| Workload | Tokens in / out | Cost |
|---|---|---|
| One long chat turn | 100K / 10K | $0.089 |
| A codebase review | 1,000K / 100K | $0.889 |
| A day of agent work | 10,000K / 1,000K | $8.89 |
Computed from Mistral Large 4’s list rates above — cache discounts and batch tiers are not applied.
Mistral Large 4 benchmark scores
| Benchmark | Score |
|---|---|
| DeepSWE[1] Long-horizon coding · ±9.5 is noise |
Who ran these numbers: 0 of 1 independent; vendor self-reported (1).
- DeepSWE: Mistral's launch-chart figure for DeepSWE v1.1, quoted in the 2026-10-06 announcement. No independent DeepSWE run exists for this model yet; per this site's rules a self-reported row never drives a verdict.
Notes on the record
Launched as a public preview on 2026-10-06 (mistral.ai news): API access via Mistral Studio, weights promised for the end of October 2026 while the model finishes red-teaming with partners. No license name announced yet — "open-weight" is the entire stated licensing plan as of this page's build (2026-10-06), so the license field reads proprietary.
Pricing has two published tiers. The docs pricing page lists the standard regional rate at $0.68 input / $0.07 cached / $2.09 output per 1M tokens while Artificial Analysis's benchmark runs bill the priority tier at $1.36/$4.18 (exactly double; the launch announcement quoted this tier).
Specs per the docs model page: 1.05T total parameters, 49B active, plus a 1.6B vision encoder; training data spans 160+ languages (vendor claims). The docs state the context as 1M tokens; Artificial Analysis's summary separately lists 524k — both are published figures and this page records the vendor's. No knowledge cutoff is disclosed anywhere.
The launch chart, all self-reported: DeepSWE v1.1 at 61.7% (the one tracked-benchmark claim, filed as a self-reported row below); a combined "Coding Agent Index" of 49.8% that Mistral says leads DeepSeek V4 Pro 0813 and Qwen3.8-Max; AutomationBench at 59.9% ahead of Kimi K3 and MiMo-V2.6-Pro; 82% on an Artificial Analysis Cyber Index vulnerability-reproduction test, which the announcement calls the highest of any model. A vendor-commissioned blind human eval (Surge AI, 1–5 scale) scored Mistral Large 4 Preview 3.74, second of five behind Claude Opus 5 at 4.22 and ahead of Kimi K3 at 3.59 — a claim about preference, not a measurement this site can reproduce.
Independent reads as of 2026-10-06: Artificial Analysis runs it in the Intelligence Index at 38, #64 of 225 (median 26), lists the model as proprietary pending the weights, and separately lists a 524k context window. No vals.ai, LiveBench, ARC Prize, or official-board rows exist yet. This is the first Mistral model tracked on this site.
Compare with
FAQ
What is Mistral Large 4?
Mistral AI's flagship: a 1.05T-parameter (49B active) natively multimodal mixture-of-experts model that entered public preview on 2026-10-06, serving a vendor-stated 1M-token context at $0.68/$2.09 per million tokens. Open weights are promised for late October 2026; until then it is API-only.
How much does Mistral Large 4 cost?
The standard regional rate is $0.68 per 1M input tokens and $2.09 per 1M output tokens, with cached input at $0.07 (docs.mistral.ai pricing page). Artificial Analysis's benchmark runs bill the priority tier at $1.36/$4.18 — exactly double — which is also the rate Mistral's own launch announcement quoted.
Has Mistral Large 4 been independently benchmarked?
Partially. Artificial Analysis runs it in the Intelligence Index at 38 (#64 of 225 models, median 26) and this page cites that figure — but none of this site's twelve tracked benchmarks carries an independent Mistral Large 4 row as of 2026-10-06. The DeepSWE v1.1 61.7% recorded here is Mistral's own number.
Is Mistral Large 4 available?
The API is live: Mistral Large 4 entered public preview on 2026-10-06 via Mistral Studio. The open weights are not out yet — Mistral promises them for late October 2026, after partner red-teaming finishes.
What context window does Mistral Large 4 have?
Mistral's docs state the context as 1M tokens. Artificial Analysis's comparison summary separately lists a 524k window for the model — both figures are published; this page records the vendor's spec and notes the difference.
Is Mistral Large 4 open source?
Not yet. The launch announcement commits to open weights by the end of October 2026, but as of 2026-10-06 the weights are not downloadable, no license name has been announced, and the model is API-only — which is why this page's license field reads proprietary.
How does Mistral Large 4 compare with other frontier models?
Everything on the launch chart is Mistral's own claim: a vendor-commissioned Surge AI blind eval scored it 3.74/5, second behind Claude Opus 5 (4.22) and ahead of Kimi K3 (3.59) and GLM-5.3 (3.60); the launch chart claims DeepSWE v1.1 61.7% and leads over DeepSeek V4 Pro and Qwen3.8-Max on coding-agent benchmarks. Zero independent scores on this site's tracked benchmarks exist yet.
What is Mistral Large 4's knowledge cutoff?
Not disclosed. Neither the launch announcement nor the docs model page states a training-data cutoff date, so this page leaves the field blank rather than borrowing an estimate.
Further reading
- Models with 10M token context windows 2026 — Mistral Large 4 is one of the 36 models it compares.
- Self-reported vs. independent AI benchmarks — the self-reported-versus-independent rule every score on Mistral Large 4’s page follows.
Benchmark guides
- DeepSWE — what Mistral Large 4’s coding score on it does and doesn’t prove.