feat: add GPT-6 family support (luna, sol, astra, 6.1-sol) - #582
Open
Siarhei-Yarkavy wants to merge 8 commits into
Open
Siarhei-Yarkavy wants to merge 8 commits into
Siarhei-Yarkavy wants to merge 8 commits into
Conversation
8nevil8
reviewed
Sep 28, 2026
Siarhei-Yarkavy
added a commit
to Siarhei-Yarkavy/codemie-code
that referenced
this pull request
Sep 28, 2026
Address review on codemie-ai#582: gpt-6-luna/sol now use GPT_6_EFFORTS instead of GPT_5_6_EFFORTS. Same values, isolated for future divergence.
This was referenced Oct 4, 2026
codemie-opencode overrode models.dev limits with heuristic defaults that had no gpt-6 branch, showing ~128k instead of the vendor-published 1050000. Refs codemie-ai#581
Address review on codemie-ai#582: gpt-6-luna/sol now use GPT_6_EFFORTS instead of GPT_5_6_EFFORTS. Same values, isolated for future divergence.
Replace startsWith('gpt-6') || /gpt-6[.-]/ with /gpt-6(?!\d)/ so bare vendor-prefixed ids (openai.gpt-6) and dotted variants (gpt-6.1-sol, future 6.2+) route via Responses API with 1050000/128000 limits, while gpt-60-style ids stay out.
Siarhei-Yarkavy
force-pushed
the
fix/gpt-6-luna-context-window
branch
from
October 4, 2026 18:21
1096df7 to
bb3fe16
Compare
| maxOutputTokens: 128000, | ||
| }, | ||
| { | ||
| family: 'gpt-6-sol', |
Collaborator
There was a problem hiding this comment.
Only gpt-6-luna and gpt-6-sol were added here, but the family has a third model, gpt-6-astra, already priced in pricing.json. Compare with gpt-5.6, where all three variants (luna/sol/terra) were added to this same table together - gpt-6-astra is missing the same treatment, so it would hit the fallback (128000 instead of 922000) if selectable via VS Code BYOK.
Author
There was a problem hiding this comment.
Fixed — gpt-6-astra got the same treatment as the gpt-5.6 trio: added to VS_CODE_CAPABILITY_TABLE (922000/128000, Responses, GPT_6_EFFORTS), to static OPENCODE_MODEL_CONFIGS, and to the test fixtures. No more 128000 fallback. gpt-6.1-sol (shipped, no astra variant) got an entry too.
Also corrected pricing while at it, sourced from the vendor Standard table (short context) — https://developers.openai.com/api/docs/pricing?latest-pricing=standard
- added the missing
gpt-6-1-solrow: $2.00 / $0.10 / $2.50 / $10.00 - fixed
gpt-6-astracacheWrite 10 → 12.5 (vendor says $12.50; 10 was a mirrored-input error carried over from feat(analytics)!: unify always-costed analytics export and strict model pricing #583)
gpt-6-sol / gpt-6-luna rows already matched the vendor table, left untouched.
VS Code capability table: gpt-6-astra entry (922000/128000, Responses, GPT_6_EFFORTS) alongside luna/sol. Addresses review thread r4183491171 in codemie-ai#582.
VS Code capability table: gpt-6.1-sol entry (922000/128000, Responses, GPT_6_EFFORTS); dotted family key per gpt-5.6 precedent, resolves dashed, vendor-prefixed and dated ids. Static OpenCode config: 1050000/128000; cost and dates mirrored from gpt-6-sol pending vendor pricing verification (no pricing row for 6.1-sol yet; no 6.1-astra shipped).
Static OpenCode config cache_read 0.20 -> 0.1; add gpt-6-1-sol row to pricing.json (2/10/0.1/2.5, dashed key per file convention; dotted and dated ids resolve via normalizeModelName/snapshot). Source: https://developers.openai.com/api/docs/pricing?latest-pricing=standard (Standard pricing table, row 'gpt-6.1-sol'). Supersedes the mirrored-from-gpt-6-sol placeholder in c751f5c.
cacheWrite 10 -> 12.5 per https://developers.openai.com/api/docs/pricing?latest-pricing=standard (Standard pricing table, row 'gpt-6-astra': short-context 0.00/.00/2.50/0.00). 10 was a mirrored-input error carried over from codemie-ai#583; real price is ~1.25x input like the other GPT-6 rows.
Siarhei-Yarkavy
changed the title
This branch has not been deployed
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.
Summary
Full GPT-6 family support: luna, sol, astra and 6.1-sol report vendor-published limits instead of heuristic fallbacks.
Closes #581.
Changes
Pricing source
Short-context Standard rows verified against https://developers.openai.com/api/docs/pricing?latest-pricing=standard (Standard pricing table).
Impact
Before: gpt-6 models showed ~128k context / fallback capabilities via codemie-opencode and VS Code BYOK. After: vendor-published windows (1050000 in opencode/Pi, 922000 VS Code input), matching standalone opencode (models.dev).
Checklist