Mistral AI Pricing
All plans and pricing as of 2026-11-05
Le Chat (Free)
- ✓Web chat interface with Mistral models
- ✓Mistral Small 4 + Medium 3.5 available (Medium 3 retired 2026-08-31)
- ✓Basic features, limited rate
API (Mistral Small 4)
- ✓119B MoE, Apache 2.0 open-weight
- ✓Unifies Small/Magistral/Pixtral/Devstral into one model
- ✓Fast, efficient, 128K context
API (Mistral Medium 3.5)
- ✓Public preview SHIPPED 2026-04-29 -- Mistral's first 'flagship merged' model
- ✓128B dense, 256k context, 77.6% SWE-Bench Verified
- ✓Underlies new Vibe Remote Agents + Le Chat Work Mode
API (Mistral Medium 3 -- RETIRED 2026-08-31)
- ✓Launched April 9, 2026
- ✓EU AI Act compliance metadata
- ✓**RETIRED 2026-08-31 alongside Medium 3.1** -- `mistral-medium-2505` was deprecated 2026-05-22 and reached retirement today; the vendor-listed replacement is **Mistral Medium 3.5**
API (Mistral Large 4, open weights since 2026-10-29)
- ✓Price cut on 2026-10-29 from the preview rate of $1.36 / $0.14 / $4.18; the docs model page still labels mistral-large-4 (v26.10) Public Preview
- ✓About 1T total parameters, 49B active per token (52B with embeddings and output layers), 128 routed experts with 4 active, 1.6B vision encoder, 1M context
- ✓Three reasoning efforts: low, medium and high (default)
- ✓Weights on Hugging Face (mistralai/Mistral-Large-4-1T-A52B, plus an NVFP4 checkpoint) under the Mistral Ecosystem Open-Weight License: modified MIT, free for commercial use, but a Model-as-a-Service business needs a separate commercial licence
- ✓Also on OpenRouter, Microsoft Foundry and NVIDIA-hosted endpoints / NIM
API (Mistral Large 3)
- ✓Flagship sparse MoE
- ✓256K context
- ✓MRL license (paid for commercial self-hosting)
API (Voxtral 3, 2026-10-20)
- ✓Voxtral Small Transcribe 3: diarization, context biasing and end-of-turn detection across 50 languages (vendor claims it beats Whisper, Scribe v2 and Gemini 3.5 Transcribe)
- ✓Voxtral Mini Transcribe 3: 4B parameters at $0.0035 per minute
- ✓Voxtral Mini TTS 3: 35+ languages and 50+ dialects, voices that adapt emotion and pace to context
- ✓Voice agent SDK (listen, reason, call tools, speak in one streaming pipeline) and Pro Voice adaptation from 30 minutes of audio
- ✓Mistral API on European, US and global endpoints, or self-deployment
Voxtral TTS
- ✓4B-param open-source speech model, March 2026
- ✓9 languages, runs on consumer hardware
- ✓Apache 2.0
Is Mistral AI Worth the Price?
Value Score: 9/10
Overall Score: 7.5/10 · Developers who want cheap, high-quality API access. Also strong for multilingual applications and European companies that prefer an EU-based AI provider for data residency.
Mistral is the scrappy underdog that keeps surprising people. Their models are impressively efficient -- you get frontier-adjacent quality at a fraction of the API cost, and Mistral Large 4 (public preview October 6, 2026; open weights under a modified MIT licence and a halved API price on October 29) puts a 1T-parameter flagship on that same cheap-and-open footing. But the consumer experience (Le Chat) is rough. This is primarily a developer's tool. If you're building AI applications on a budget, Mistral should be on your shortlist.
The Tier List Tuesday
Weekly newsletter: tier movers, new entrants, and the VS of the week. Built from our daily AI-tool sweeps. No spam, unsubscribe anytime.
How Mistral AI Pricing Compares
| Tool | Free Tier | Starting Price | Value Score | Overall |
|---|---|---|---|---|
| Mistral AI(this tool) | Yes | $0 | 9/10 | 7.5 |
| Qwen (Alibaba) | Yes | $0 | 10/10 | 8.8 |
| MiniMax M3 | Yes | $0 | 9.5/10 | 8.4 |
| Gemma 4 (Google) | Yes | $0 | 10/10 | 8.3 |
| IBM Granite 4.2 | Yes | $0 | 9.5/10 | 8.2 |
| Kimi K3.1 Pro / K3 (Moonshot) | Yes | $2 / $10/per 1M tokens (input/output) | 8.5/10 | 8.1 |