Model comparison
Mistral Medium 3.5 vs GPT-6 Sol: 32% less per session
Mistral Medium 3.5 undercuts GPT-6 Sol by 25% per token and 32% on a cached coding session. The trade-offs: a 256K window, preview status, and fewer tools.
· Prices as of September 28, 2026
Mistral Medium 3.5
Mistral AI · Released May 22, 2026 · Preview
Mistral's coding and agent flagship, a dense open-weight model that replaced Devstral 2 in Mistral's Vibe coding agents.
Mistral Medium 3.5 facts and comparisonsGPT-6 Sol
OpenAI · Released September 22, 2026
The mid-priced GPT-6 model, which OpenAI pitches for complex coding and agent workflows and which the Codex docs recommend for complex coding.
GPT-6 Sol facts and comparisons
The short answer
Mistral Medium 3.5 costs $1.43 on the example agentic coding session against $2.10 on GPT-6 Sol, 32% less, from rates 25% lower on input, output, and cache hits plus the absence of a cache-write premium. GPT-6 Sol brings a 1.05M context window, Codex, and GitHub Copilot, and the Codex docs recommend it for complex coding. Mistral Medium 3.5 is a public preview with a 256K window and open weights, reachable here through OpenRouter and used in Mistral's Vibe agents.
Choose Mistral Medium 3.5 if
- Cost leads the decision: $1.50 input and $7.50 output per million tokens, against $2 and $10.
- Your requests fit in 256K tokens and you want written cache tokens billed at plain input, not OpenAI's 1.25x.
- You want open weights: Mistral releases Medium 3.5 under a modified MIT license, and says four GPUs can host the dense model.
Choose GPT-6 Sol if
- You live in Codex, where the docs name GPT-6 Sol as the pick for complex coding.
- Some requests need more than 256K tokens: GPT-6 Sol's window is 1.05M, with room for up to 922K of input.
- You want a model that is generally available rather than in public preview.
- Your team uses OpenCode or GitHub Copilot, which list GPT-6 Sol and not Mistral Medium 3.5.
Side by side
Specs and prices
| Fact | Mistral Medium 3.5 | GPT-6 Sol |
|---|---|---|
| Maker | Mistral AI | OpenAI |
| API model id | mistral-medium-3.5 | gpt-6-sol |
| Released | May 22, 2026 | September 22, 2026 |
| Status | Preview | Current |
| Context window | 256K tokens | 1.05M tokens |
| Max output | Not published | 128K tokens |
| Open weights | Yes | No |
| Input, per 1M tokens | $1.50 | $2 |
| Cache hit, per 1M | $0.15 | $0.20 |
| Cache write, per 1M | $1.50 (same as input) | $2.50 |
| Output, per 1M tokens | $7.50 | $10 |
| Runs in | OpenRouter | Codex, OpenCode, OpenRouter, and GitHub Copilot |
Standard API rates in US dollars, as published by each maker on September 28, 2026. Batch and priority tiers, taxes, and subscription plans are not included. Mistral Medium 3.5: Mistral lists no per-model cache price. Its caching docs bill cached tokens at 10% of input, which is the $0.15 shown. GPT-6 Sol: Requests over 272K input tokens cost 2x for input and cache and 1.5x for output, for the whole request.
Cost
What the same work costs
The same token counts, priced at each model’s published rates. Cached tokens are billed at each maker’s cache prices, so the session shows what caching is worth on each model.
Real sessions differ: the two models count the same code as different numbers of tokens, and reasoning settings change how much each one writes. Your own history is the real test.
| Workload | Mistral Medium 3.5 | GPT-6 Sol |
|---|---|---|
| Agentic coding session, 100K input, 400K written to cache, 2M read from cache, 50K output | $1.43 | $2.10 |
| Large one-off review, 150K input with no cache hits, 10K output | $0.30 | $0.40 |
| Output-heavy generation, 30K input, 80K output | $0.65 | $0.86 |
| A month of sessions, 110 sessions: 5 a day, 22 working days | $156.75 | $231.00 |
| Where the session’s cost goes | ||
| Cache writes | $0.60 | $1.00 |
| Cache reads | $0.30 | $0.40 |
| Uncached input | $0.15 | $0.20 |
| Output | $0.38 | $0.50 |
- caching saves on the session with Mistral Medium 3.5 (65%)
- $2.70
- caching saves on the session with GPT-6 Sol (62%)
- $3.40
Where the 32% saving comes from
Mistral Medium 3.5 lists $1.50 per million input tokens and $7.50 per million output tokens, while GPT-6 Sol lists $2 and $10. Cache hits keep the same proportion, since Mistral's caching rule bills cached tokens at 10% of input, $0.15, and OpenAI charges $0.20. On anything uncached the difference is a flat 25%, which is why the large one-off review costs $0.30 against $0.40 and the output-heavy generation $0.65 against $0.86.
Writes supply the rest. From GPT-5.6 on, OpenAI bills a cache write at 1.25x input, $2.50 per million on GPT-6 Sol, and Mistral charges nothing extra, so its writes cost the $1.50 input rate. The example session's writes therefore come to $0.60 against $1.00, and the whole session to $1.43 against $2.10. Across 110 sessions a month that is $156.75 against $231.00, $74.25 apart.
Any prompt Mistral Medium 3.5 accepts stays under GPT-6 Sol's 272K line
GPT-6 Sol carries a long-context rule: once a request passes 272K input tokens, input and cache cost 2x and output 1.5x, for the entire request. Mistral Medium 3.5's whole window is 256K, so any prompt it can take would also sit below that line on GPT-6 Sol. Mistral's price data shows no long-context tier.
Up to 256K, then, the two compare cleanly at the rates in the tables. Past that point GPT-6 Sol is the one that can take the request, with a 1.05M window and up to 922K of input, and its higher rates apply above 272K. GPT-6 Sol writes up to 128K of output per response, and Mistral publishes no output ceiling.
Preview status, open weights, and tool support
Mistral announced Medium 3.5 as a public preview. It is a dense model with weights under a modified MIT license, and it replaced Devstral 2, retired on July 30, 2026, in Mistral's Vibe coding agents. Mistral calls it "our frontier-class multimodal model optimized for agentic and coding use cases." This post does not put a price on self-hosting.
OpenAI released GPT-6 Sol on September 22, 2026, as the middle model of the GPT-6 family, "built to power complex coding and agentic workflows." It runs in Codex, OpenAI's own coding agent, with reasoning effort at medium by default, and in OpenCode, OpenRouter, and GitHub Copilot. Codex steers users of GPT-5.6 Sol, GPT-5.6 Terra, and GPT-5.4 toward it.
Of the tools here, Mistral Medium 3.5 is reachable through OpenRouter, which passes an open-weight model's requests to one of several providers whose prices can differ from Mistral's own API. EveryToken prices it there from OpenRouter's catalog, and prices GPT-6 Sol at OpenAI's rates from Codex and OpenCode history.
Prompt caching
How each maker bills cached tokens
Mistral AI
Mistral caches prompt prefixes in 64-token blocks. A prompt_cache_key raises the chance of a hit but doesn't guarantee one.
Cached prompt tokens are billed at 10% of the standard input price, and Mistral lists no fee for writing the cache.
Source: Mistral docs: Prompt caching
OpenAI
Prompt caching is on by default. From GPT-5.6 on you can also mark up to four explicit cache breakpoints, while GPT-5.5 and earlier cache automatically only.
From GPT-5.6 on, a cache write costs 1.25x the uncached input price and a cache hit costs 0.1x. GPT-5.5 and earlier add no charge for writing the cache: written tokens are billed as ordinary input, and a hit costs 0.1x on the models compared here.
On GPT-5.6 and later, a cached prefix stays reusable for at least 30 minutes after its last use, and caching starts at 1,024 input tokens.
Source: OpenAI: Prompt caching
Your own numbers
See what Mistral Medium 3.5 and GPT-6 Sol really cost you.
everyaitoken reads your OpenRouter, Codex, and OpenCode history on your Mac and prices every request at API rates, with what caching saved or cost. $9 once.
FAQ
Questions
How much cheaper is Mistral Medium 3.5 than GPT-6 Sol?
Per token, 25% on input, output, and cache hits. On the example agentic session the saving grows to 32%, $1.43 against $2.10, because GPT-6 Sol bills cache writes at 1.25x input.
Does GPT-6 Sol charge more to write the cache?
Yes. OpenAI bills cache writes from GPT-5.6 on at 1.25x the input price, $2.50 per million on GPT-6 Sol. Mistral lists no fee for writing the cache, so Mistral Medium 3.5 bills written tokens at $1.50.
What context window does each model have?
Mistral Medium 3.5 accepts 256K tokens. GPT-6 Sol has a 1.05M window with up to 922K of input, and bills requests over 272K input tokens at higher rates.
Where can I use Mistral Medium 3.5?
Among the tools compared here, through OpenRouter, and Mistral uses it to power its Vibe remote coding agents. GPT-6 Sol runs in Codex, OpenCode, OpenRouter, and GitHub Copilot.
Sources
- Mistral docs: Mistral Medium 3.5
- Mistral: Vibe remote agents and Mistral Medium 3.5
- Mistral docs: Models
- OpenRouter: Mistral Medium 3.5
- OpenAI: API pricing
- OpenAI docs: GPT-6 Sol
- OpenAI: Introducing GPT-6 Sol and Luna
- OpenAI: API changelog
- Codex docs: Models
- OpenRouter: GPT-6 Sol
- OpenCode docs: Zen
- GitHub Docs: Supported AI models in Copilot
- Mistral docs: Prompt caching
- OpenAI: Prompt caching