How to Use Mistral Medium 3.5 in Codex
Mistral Medium 3.5 handles multi-step coding, research, and structured extraction with a 256k context window. It combines instruction following and reasoning, and outscored Devstral 2 on Mistral's SWE-Bench Verified evaluation.
Prompts to try with Mistral Medium 3.5 in Codex
Copy an example into Codex, or ask in your own words.
Use Mistral Medium 3.5 with high reasoning effort to investigate the failing tests in this project. Reproduce the failure, trace the cause, make the smallest justified patch, and run the relevant tests.
Use Mistral Medium 3.5 to review the current diff for behavior changes. Return at most five findings, each with a file location, a concrete failure scenario, and a suggested check.
Use Mistral Medium 3.5 with high reasoning effort to refactor duplicated validation code in this repository. Preserve public behavior, explain the proposed boundary, implement it, and run the affected tests.
Use Mistral Medium 3.5 to compare this project's API implementation with its specification. List mismatches with code references, then add tests for confirmed gaps.
Use Mistral Medium 3.5 to investigate the latest CI failure using the available logs and project files. Separate observed errors from hypotheses, test the leading explanation, and report the result in under 300 words.
How to set up Mistral Medium 3.5 in Codex
Pick the way that suits you. Each takes about two minutes.
Use Spiral's MCP server
- Add the hosted server
codex mcp add spiral --url https://api.spiral.ai/mcp
- Sign in
Codex opens the browser: sign in to Spiral and click Authorize. No API key. codex mcp list then shows Spiral.
codex mcp login spiral
Use the Spiral skill
- Send Codex this message
Your agent reads Spiral's skill and sets itself up.
set up https://spiral.ai/SKILL.md
- Sign in or paste a key
Approve Spiral in the browser, or give the agent a key from the console.
Use the Spiral CLI
- Install and sign in
Installs the Spiral CLI, signs you in and adds its skill to your coding agents.
npx -y spiral-cli@latest init
Call the Spiral API
- Create an API key
In the console, under API keys.
- Run Mistral Medium 3.5
One request; the result comes back by polling or a webhook.
curl https://api.spiral.ai/v1/runs \ -H "Authorization: Bearer $SPIRAL_API_KEY" \ -H "Content-Type: application/json" \ -H "Idempotency-Key: $(uuidgen)" \ -d '{ "provider": "openrouter", "model": "mistralai/mistral-medium-3-5", "input": { "messages": [ { "role": "system" } ] } }'
Then paste one of the prompts above: Codex finds Mistral Medium 3.5 on Spiral, shows the price and runs it.
Mistral Medium 3.5 providers and prices
Spiral runs Mistral Medium 3.5 through OpenRouter. Open a card for today's price, with every fee included.
Up to $0.040005 per run at the default inputs. $1.95 per 1M input tokens, $9.75 per 1M output tokens.
Settings you can ask for
Ask Codex for any of these in your prompt.
| Setting | Allowed values | Default | Description |
|---|---|---|---|
| Seed | a whole number | ||
| Top p | 0 to 1 | ||
| Max tokens | 1 to 209715 | The most output tokens to generate. | |
| Temperature | 0 to 2 | ||
| Presence penalty | -2 to 2 | ||
| Reasoning effort | none or high | high | How much the model reasons. |
| Frequency penalty | -2 to 2 |
Every API field of Mistral: Mistral Medium 3.5 is on its tool page: open its card above.
What can you do with Mistral Medium 3.5 in Codex?
Turn a bug report or refactoring request into coordinated edits across project files with Mistral Medium 3.5. Give the agent a reproduction, behavior constraints, and tests that define completion.
Bug fixes across several modules
Trace a reported failure through related code, propose a patch, and check it against a reproduction and regression tests.
Module refactoring
Refactor a module against explicit behavior requirements, then summarize the changes and test results for review.
Research briefs with evidence
Combine supplied research into a brief that links conclusions to sources and separates open questions from supported findings.
Structured record extraction
Turn supplied text into consistent records using a defined schema, examples, and a rule for missing values.
Multilingual support triage
Classify English, French, Spanish, and German inquiries, summarize the issue in English, and draft a reply in the customer's language.
Recurring operational digests
Summarize incoming updates into decisions, unresolved issues, and next actions, preserving the source of each item.
Long-document comparison
Compare supplied reports or specifications within the 256k context window and produce a table of agreements, contradictions, and unanswered questions.
Test planning and generation
Convert requirements and implementation details into tests covering expected behavior, boundary conditions, and failure cases.
More on Mistral Medium 3.5 and Codex
Overview
Mistral Medium 3.5 in other agents
Questions about Mistral Medium 3.5 in Codex
Can Codex use Mistral Medium 3.5?
Yes. Add Spiral to Codex as an MCP server, and Codex can find and run Mistral Medium 3.5 like any built-in tool.
How much does Mistral Medium 3.5 cost in Codex?
Up to $0.040005 per run at the default inputs. $1.95 per 1M input tokens, $9.75 per 1M output tokens. You pay from your Spiral balance as you go, with no subscription, and a failed run is not charged. Your Codex plan is separate.
Do I need an OpenRouter account to use Mistral Medium 3.5 in Codex?
No. Spiral runs Mistral Medium 3.5 on its own OpenRouter account. You sign in to Spiral once and never paste a key.
How long does it take to set up Mistral Medium 3.5 in Codex?
About two minutes: add the hosted server, then sign in.
Does Codex wait while Mistral Medium 3.5 runs?
Mistral Medium 3.5 runs in the background. Codex starts the run, checks on it and gives you the text, json when it's ready.
What is Mistral Medium 3.5 best suited for?
Multi-step coding, research synthesis, and tasks that combine tool use with structured output. Give it a concrete deliverable, such as a tested patch, an evidence table, or extracted records.
How much context can Mistral Medium 3.5 handle?
Mistral specifies a 256k-token context window. Treat that as a context limit, not a promise that it can write a 256k-token answer.
When should I enable reasoning?
Mistral recommends reasoning for complex prompts and agentic coding. Its recommended temperature with high reasoning effort is 0.7.
Can Mistral Medium 3.5 work in multiple languages?
Yes. Supported languages include English, French, Spanish, German, Russian, Chinese, and Japanese; specify the language you want in the result.
What are Mistral Medium 3.5's weaknesses?
NVIDIA documents weaker non-agentic benchmark results and more verbose responses than some closed-source competitors. Ask for a fixed length and explicit acceptance criteria, and check generated code with tests.
Start using Mistral Medium 3.5 in Codex
One Spiral account for every tool your agent needs. No subscription; you pay per run.
Paste this into Codex:
set up https://spiral.ai/SKILL.md


