Mistral Medium 3.5 API providers and pricing evidence

Route-specific buying guide

How to budget Mistral Medium 3.5 on OpenRouter

Route ID: mistralai/mistral-medium-3.5-20260430

The accepted rows describe Mistral Medium 3.5 on Mistral's owner-direct API. OpenRouter is shown as a separate access channel, but its reviewed endpoint identifies Mistral as the upstream provider; two channels do not prove infrastructure diversity or a market-wide cheapest provider.

Mistral's exact model card reports a 256,000-token context window for this release. Context capacity is a technical ceiling, not a budget target, and the provider's live limits, account tier, and endpoint behavior still need verification before purchase.

Cost formula

Divide uncached input, cache-read input, and output token counts by one million, then multiply each by its reviewed meter. If a channel lacks a reviewed cache meter, exclude that channel whenever cached-input usage is above zero instead of borrowing the ordinary input rate.

Costs that move the bill

• Prompt and retrieved-context size across the actual request distribution
• Cache-hit ratio and whether the selected access channel exposes a reviewed cache-read meter
• Completion length, tool loops, retries, failed calls, and fallback routing
• Quotas, context support, latency, regional access, data terms, and support

Checks before purchase

• Pin mistral-medium-3-5-26-04 or the reviewed dated OpenRouter alias in configuration and logs.
• Confirm that the route has not moved to a newer alias before applying these meters.
• Reconcile uncached, cached, and output usage fields against invoice samples.
• Test failure, timeout, and retry behavior without assuming the two channels provide independent upstream capacity.

Source checked Aug 23, 2026: OpenRouter public model API. Recheck the live route before committing spend.

Estimate this route with the LLM workload calculator →

Compare another model family

• Qwen API pricing
• Kimi API pricing
• GPT-OSS 120B pricing

Decision guide

How to compare Mistral Medium 3.5 API providers

Mistral Medium 3.5 is a versioned service release, not a generic Mistral family label. APIDir maps the owner-direct model ID mistral-medium-3-5-26-04 to OpenRouter's dated mistralai/mistral-medium-3.5-20260430 route only because Mistral's model card and lifecycle documentation identify the release boundary. A future alias or successor must be reviewed as a new identity instead of silently inheriting this price table.

The current evidence covers Mistral's direct API and an OpenRouter access channel. Both publish the same standard input and output token rates, but OpenRouter's endpoint inventory identifies Mistral as the upstream provider. These are two ways to buy access, not proof of two independent inference infrastructures. APIDir therefore shows both costs without calling either route the market winner.

Cache use changes the eligible sample. Mistral publishes a separate cached-input rate for the owner-direct service, while the reviewed OpenRouter response does not expose an accepted cached-input meter for this route. When a workload contains cached tokens, the router row is removed rather than charged at the ordinary input rate. The result becomes one verified price card and remains explicitly unranked.

What changes Mistral Medium 3.5 API cost

• Uncached input, cache-read input, output length, and the workload's real cache-hit ratio
• Owner-direct versus router access, route fallback policy, quotas, context support, and endpoint compatibility
• Retries, tool loops, failed requests, storage, networking, taxes, data terms, and production support

Integration checks before production

• Pin mistral-medium-3-5-26-04 or the reviewed dated router alias and reject silent model substitution
• Treat cached-input support as a provider capability; never substitute the uncached rate for a missing meter
• Log provider route, resolved upstream when available, token meters, retry count, terminal error, and request ID

Workloads this comparison can inform

Production language applications

Estimate uncached input and output from representative request traces, then compare quota, latency, reliability, and support outside the price table.

Repeated-context workflows

Model cache reads separately and keep a no-cache fallback budget because only the owner-direct cached meter currently passes review.

Router-based procurement

Use the routed channel for access convenience only after verifying upstream visibility, fallback controls, data terms, and the exact dated model alias.

How to read the price result

Two access channels pass the exact Mistral Medium 3.5 input and output scope at the reviewed rates. APIDir publishes the comparison but no #1; cached workloads currently reduce to one owner-direct price card.

Related model price comparisons

• Qwen API pricing comparisonCompare Qwen API pricing by exact model ID, provider channel, input and output token meter, context, and official source.
• Kimi API pricing comparisonCompare Kimi API pricing across provider routes without mixing Kimi versions, token meters, context tiers, or service channels.
• GPT-OSS 120B pricing comparisonCompare source-linked GPT-OSS 120B input and output token prices across provider routes while keeping infrastructure compute meters separate.