Mistral Large 3 is far cheaper than Claude Sonnet 5.5 per token, but this site lists no cache pricing for it, its window is a quarter the size, and Mistral has deprecated it in favour of Mistral Large 4.
Mistral Large 3 is 4× cheaper on input and about 6.7× on output than Claude Sonnet 5.5. On cache-heavy agent work the gap almost disappears, because Sonnet 5.5 bills cached reads at $0.10. Mistral deprecated Large 3 on October 31, 2026 (retirement April 30, 2027) and recommends Mistral Large 4 for new integrations.
| Mistral Large 3 | Claude Sonnet 5.5 | |
|---|---|---|
| Provider | Mistral | Anthropic |
| Tier | Mid-tier | Mid-tier |
| Input / 1M tokens | $0.50 | $2.00 |
| Output / 1M tokens | $1.50 | $10.00 |
| Cache hit / 1M | — | $0.10 |
| Cache write / 1M | — | $2.50 |
| Context window | 256K | 1M |
| Notes | Former Mistral flagship; deprecated Oct 31, 2026, retiring Apr 30, 2027 — replaced by Mistral Large 4 | Released Sep 28, 2026, replacing Sonnet 5 at the identical $2/$10 per 1M price; Sonnet 5 is now a legacy model |
Mistral Large 3 is 4.0× cheaper on input tokens ($0.50 vs $2.00 per 1M). Mistral Large 3 is 6.7× cheaper on output tokens ($1.50 vs $10.00 per 1M).
| Workload | Mistral Large 3 | Claude Sonnet 5.5 | Difference |
|---|---|---|---|
| Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | $16.00 | $80.00 | Mistral Large 3 saves $64.00 |
| Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | $16.20 | $68.00 | Mistral Large 3 saves $51.80 |
| Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens | $2.23 | $2.66 | Mistral Large 3 saves $0.44 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
you already run it in production on uncached, short-prompt workloads and need a cheap option until its April 30, 2027 retirement — or you want Apache 2.0 open weights to self-host.
you're starting a new integration, need more than 256K tokens of context, or your workload re-reads a cached context where Sonnet's cache pricing applies.
Mistral's model page lists Large 3 (mistral-large-2512) as deprecated on October 31, 2026 and says to use Mistral Large 4 for new integrations; retirement is scheduled for April 30, 2027. Building on Large 3 now means planning a migration within six months. Large 3 is an open-weight model under Apache 2.0, so self-hosting remains possible after API retirement.
This site has no cache-hit price for Mistral Large 3, so the coding-agent example bills all its input at $0.50. Sonnet 5.5 bills cache hits at 5% of input. On a session that is 90% cached, the two cost nearly the same; on uncached chat, Mistral is several times cheaper.
Large 3's window is 256K; Sonnet 5.5's is 1M, at standard rates throughout. Documents over 256K tokens need chunking on Mistral. Both providers advertise 50% off for batch processing.
Per token, yes: $0.50 vs $2.00 input and $1.50 vs $10.00 output per 1M. On heavily cached workloads the totals are close.
Yes. Mistral lists it as deprecated on October 31, 2026 with retirement on April 30, 2027, and recommends Mistral Large 4.
Yes. Mistral describes it as an open-weight model under the Apache 2.0 licence.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.