Add MiniMax-M3 model (#1933)#1940
Merged
Merged
Conversation
MiniMax-M3 is MiniMax's new frontier multimodal coding model: 1M context window (512K minimum on ollama-cloud), native text/image/video input, tool calling, reasoning, and open weights. Adds the model to all five providers where it should be available: - minimax (pay-as-you-go) - minimax-cn (pay-as-you-go, China) - minimax-coding-plan (token plan subscription) - minimax-cn-coding-plan (token plan subscription, China) - ollama-cloud Closes anomalyco#1933. Notes for reviewers: - Cost fields on minimax/minimax-cn match M2.7; M3 docs state the pricing is unchanged from M2.7. - The 1M/512K context divergence on ollama-cloud is intentional — ollama advertises 1M with a 512K minimum, so 512K is the safe floor that won't surprise opencode users with mid-request rejections. - output = 131072 is inherited from the existing M2.7 files; MiniMax's published M3 docs only advertise the 1M input context, not a separate output cap. - The minimax-coding-plan variant has been verified end-to-end in opencode against the MiniMax token plan API.
rekram1-node
force-pushed
the
add-minimax-m3
branch
from
3a863d3 to
f33ff9b
Compare
Collaborator
|
MiniMax is currently limiting the context window to 500k temporarily, so this PR caps the MiniMax-M3 entries at 500k for now. |
This was referenced Jun 2, 2026
|
thank you |
rekram1-node
pushed a commit
that referenced
this pull request
Jun 2, 2026
The M3 entries added in #1940 copied M2.7's cost values. Correct them to the official M3 pricing and limits: - minimax / minimax-cn (pay-as-you-go): input 0.30 -> 0.60, output 1.20 -> 2.40, cache_read 0.06 -> 0.12, and remove cache_write (M3 has no active prompt-cache-write tier). - max output 131072 -> 128000 across all four providers. - coding-plan variants keep their subscription-plan zero pricing; only max output is corrected. Context (512K), modalities, and the other flags are unchanged.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds the MiniMax M3 model to the five providers where it should be available.
What this PR does
MiniMax-M3 is MiniMax's new frontier multimodal coding model: 1M context window, native text/image/video input, tool calling, reasoning, and open weights.
Adds
MiniMax-M3.toml(andminimax-m3.tomlfor ollama-cloud) to:providers/minimax/— pay-as-you-go, globalproviders/minimax-cn/— pay-as-you-go, Chinaproviders/minimax-coding-plan/— token plan subscription, globalproviders/minimax-cn-coding-plan/— token plan subscription, Chinaproviders/ollama-cloud/— ollama cloud,minimax-m3:cloudSource
Closes #1933.
Notes for reviewers
minimaxandminimax-cnmatches M2.7. MiniMax's own docs state the M3 pricing is unchanged from M2.7.ollama-cloudis intentional. ollama advertises M3 as having a 1M context window with a 512K minimum, so 512K is the safe floor — using 1M there would mean opencode users could hit silent mid-request rejections.output = 131072on the limit block is inherited from the existing M2.7 files in this repo. MiniMax's published M3 docs only advertise the 1M input context window, not a separate output cap. Worth verifying against the M3 changelog when reviewers have access.minimax-coding-planvariant verified end-to-end in opencode against the MiniMax token plan API (auth, streaming, tool calls, reasoning) by the author before opening this PR.Validation
bun validateerrors on a pre-existing issue (Unable to resolve extends.from: mistral/mistral-medium-latest) that also occurs on a cleandevcheckout — not introduced by this changeset. CI may need to be re-run once that's fixed upstream.The five new TOML files parse cleanly and conform to the schema in
packages/core/src/schema.ts.