fix(github-copilot): claude-opus-4.8 context window to 1M with reasoning support by yichu10c · Pull Request #91875 · openclaw/openclaw

yichu10c · 2026-06-10T07:12:45Z

Summary

Updates model metadata in the GitHub Copilot plugin to match live Copilot endpoint capabilities.

Changes

Added to with:
- (was 128,000 — the default)
- (was 8,192)

Updated the bundled manifest entry for with the same values.

Why

Copilot's live endpoint returns and for . The bundled catalog was mirroring metadata which carries . This caused:

Session context budgets capped at 128K (visible in → )
No adaptive thinking/reasoning support despite upstream supporting it

Fixes #91869.

Test Results

✓ extensions/github-copilot/models.test.ts (38 tests) — 64ms
✓ extensions/github-copilot/index.test.ts (16 tests) — 1.93s

…ing support Updates claude-opus-4.8 model metadata to match live Copilot /models endpoint capabilities: - contextWindow: 128000 → 1000000 (native 1M on Anthropic API) - maxTokens: 8192 → 64000 - reasoning: false → true - input: ["text", "image"] (already set, made explicit) - compat.supportedReasoningEfforts: ["low","medium","high","xhigh","max"] Fixes openclaw#91869

clawsweeper · 2026-06-10T07:15:26Z

Codex review: needs real behavior proof before merge. Reviewed June 15, 2026, 10:06 AM ET / 14:06 UTC.

Summary
The PR updates GitHub Copilot claude-opus-4.8 static override and bundled manifest metadata to advertise 1M context, 64k output, reasoning, image input, and low/medium/high/xhigh/max reasoning efforts.

PR surface: Source +16. Total +16 across 2 files.

Reproducibility: yes. at source level, but not end to end. Current main and v2026.6.6 still carry the 128k/8192/no-reasoning fallback row, while full live reproduction of the Copilot /models row and after-fix /status behavior still needs an entitled account.

Review metrics: 1 noteworthy metric.

GitHub Copilot metadata surfaces: 1 manifest row changed, 1 static override added. These provider-owned fallback surfaces feed different catalog/runtime paths and can change upgrade behavior when live discovery is unavailable.

Stored data model
Persistent data-model change detected: vector/embedding metadata: extensions/github-copilot/model-metadata.ts. Confirm migration or upgrade compatibility proof before merge.

Merge readiness
Overall: 🧂 unranked krab
Proof: 🧂 unranked krab
Patch quality: 🐚 platinum hermit
Result: blocked until stronger real behavior proof is added.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

[P1] Add redacted live proof from an entitled Copilot setup showing the after-fix /models-derived values and OpenClaw-visible models list, /status, or thinking behavior.
Reconcile the targeted fallback correction with test(github-copilot): cover live xhigh reasoning for non-Claude mini models (#59416) #91728 so maintainers know whether one or both should land.

Proof guidance:

[P1] Needs stronger real behavior proof before merge: The PR body lists unit tests only; a later comment reports live Copilot /models values, but it does not show after-fix OpenClaw behavior from this patch. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Risk before merge

[P1] The PR still lacks after-fix real OpenClaw behavior proof from an entitled Copilot setup, so the live /models claim and /status budget improvement are not proven for this branch.
[P2] Changing shipped fallback context, max-token, and reasoning metadata for an existing provider model can affect users when live discovery is unavailable or entitlements differ.
[P2] The live-first structural PR at test(github-copilot): cover live xhigh reasoning for non-Claude mini models (#59416) #91728 may be the more durable fix for successful discovery, so maintainers should decide whether to land this targeted fallback correction, the structural fix, or both.

Maintainer options:

Prove and Reconcile the Metadata Change (recommended)
Before merge, add redacted live proof from an entitled Copilot setup showing the after-fix OpenClaw models list, /status, or thinking behavior, then decide how this targeted fallback row should coexist with the live-first catalog PR.
Land Only the Structural Live-First Path
Maintainers may choose to land test(github-copilot): cover live xhigh reasoning for non-Claude mini models (#59416) #91728 first and then reassess whether static fallback metadata still needs a separate 4.8 correction.
Accept the Static Fallback Risk
Maintainers could intentionally accept the one-model fallback bump after reviewing the external provider contract, knowing it may still drift when Copilot changes entitlements or limits again.

Next step before merge

[P2] The remaining blocker is contributor proof plus maintainer choice between this targeted fallback bump, the structural live-first PR, or both; that is not a safe automated repair lane.

Security
Cleared: The diff only changes GitHub Copilot model metadata and does not add dependencies, workflows, permissions, package resolution, downloads, or secret-handling paths.

Review details

Best possible solution:

Keep the targeted fallback metadata correction open until redacted live proof confirms the exact Copilot 4.8 row and maintainers reconcile it with the live-first catalog PR.

Do we have a high-confidence way to reproduce the issue?

Yes at source level, but not end to end. Current main and v2026.6.6 still carry the 128k/8192/no-reasoning fallback row, while full live reproduction of the Copilot /models row and after-fix /status behavior still needs an entitled account.

Is this the best way to solve the issue?

Unclear as submitted. The provider-owned manifest/static override boundary is a plausible narrow fix for fallback metadata, but the durable successful-discovery fix may be #91728 and this branch still needs after-fix proof.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 66079161d72a.

Label changes

Label justifications:

P2: This is a normal-priority GitHub Copilot provider capability fix with bounded but real model budget and reasoning impact.
merge-risk: 🚨 compatibility: The PR changes shipped fallback context, max-token, and reasoning metadata for an existing model, which can alter behavior for existing users on upgrade.
merge-risk: 🚨 auth-provider: The change affects provider model capability metadata and reasoning routing for GitHub Copilot, while the exact live entitlement contract is not proven after this patch.
rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🧂 unranked krab and patch quality is 🐚 platinum hermit.
status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs stronger real behavior proof before merge: The PR body lists unit tests only; a later comment reports live Copilot /models values, but it does not show after-fix OpenClaw behavior from this patch. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Evidence reviewed

PR surface:

Source +16. Total +16 across 2 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	2	19	3	+16
Tests	0	0	0	0
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	2	19	3	+16

What I checked:

Root policy read: Read the full root AGENTS.md and applied the provider/plugin ownership, external API proof, real behavior proof, and compatibility-risk guidance to this PR review. (AGENTS.md:1, 66079161d72a)
Scoped extension policy read: The scoped extension guide applies because bundled plugin metadata must stay accurate for discovery/setup and provider-specific catalog behavior belongs inside the plugin boundary. (extensions/AGENTS.md:1, 66079161d72a)
Current main still has stale manifest fallback metadata: Current main still declares claude-opus-4.8 with contextWindow: 128000, maxTokens: 8192, and no manifest reasoning or compat.supportedReasoningEfforts, so the central fallback metadata change is not already implemented on main. (extensions/github-copilot/openclaw.plugin.json:46, 66079161d72a)
Current main fallback builder still defaults 4.8 to 128k/no reasoning: buildCopilotModelDefinition only uses a static override when present; current main has no claude-opus-4.8 override, so the default fallback remains reasoning: false, 128k context, and 8192 max tokens. (extensions/github-copilot/models-defaults.ts:41, 66079161d72a)
Current main now wires max thinking when compat/static metadata advertises it: Current main’s Copilot thinking profile reads max from live or static compat metadata, so the previous max-profile concern is no longer a current-main blocker for this branch after merge. (extensions/github-copilot/model-metadata.ts:119, 66079161d72a)
Anthropic transport can map Opus 4.8 max without an explicit static thinkingLevelMap: resolveClaudeNativeThinkingLevelMap fills native xhigh/max mappings when no narrower model thinkingLevelMap is present, and the transport uses that map before sending adaptive thinking effort. (packages/llm-core/src/model-contracts/anthropic.ts:78, 66079161d72a)

Likely related people:

saju01: Merged PR metadata and git history show the earlier Opus 4.8 static catalog addition came through feat(github-copilot): add Claude Opus 4.8 to default model catalog #88547, which this PR adjusts. (role: introduced catalog row and adjacent contributor; confidence: high; commits: fbb776d92c29, 60580c48679a; files: extensions/github-copilot/openclaw.plugin.json, extensions/github-copilot/models-defaults.ts, extensions/github-copilot/models.test.ts)
steipete: GitHub PR metadata shows steipete merged the original Opus 4.8 catalog PR, and git history shows adjacent Copilot 1M capability work. (role: merger and recent provider metadata contributor; confidence: high; commits: fbb776d92c29, 5965522af5a7; files: extensions/github-copilot/openclaw.plugin.json, extensions/github-copilot/model-metadata.ts, extensions/github-copilot/models.ts)
Eduardo Piva: Git history shows the Copilot live /models discovery and mapping path was introduced in the same provider area. (role: live catalog discovery contributor; confidence: medium; commits: 75405f64d0a7; files: extensions/github-copilot/index.ts, extensions/github-copilot/models.ts, extensions/github-copilot/models.test.ts)
Vincent Koc: Git history shows recent Copilot thinking compatibility work adjacent to the reasoning-effort behavior affected by this PR. (role: recent thinking-compat contributor; confidence: medium; commits: 2fc5072021a2; files: extensions/github-copilot/index.ts, extensions/github-copilot/provider-policy-api.ts)

What the crustacean ranks mean

🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works

ClawSweeper keeps one durable marker-backed review comment per issue or PR.
Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
Maintainers can also comment @clawsweeper review to request a fresh review only.
Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

saju01 · 2026-06-14T16:00:21Z

Cross-link: #91728 ("prefer live model catalog") is the structural root-cause fix for this — instead of hard-coding claude-opus-4.8's window per release, it makes the live Copilot /models metadata win over stale static rows, so context window / max output / reasoning are auto-detected for every model.

Confirmed today on a real entitled account: live /models reports claude-opus-4.8 at max_context_window_tokens=1000000, max_prompt_tokens=936000, max_output_tokens=64000, supports.adaptive_thinking=true — the same 1M window the VS Code "long context" toggle exposes (not a separate model id). This manual bump fixes the symptom for one model; #91728 removes the need for these bumps going forward. Linking so maintainers can decide whether to land the targeted fix here, the structural fix in #91728, or both.

openclaw-clownfish · 2026-06-15T23:15:19Z

Thanks @yichu10c for the targeted Copilot metadata fix. I am closing this as superseded by #91728 because both PRs address the same claude-opus-4.8 live metadata shadowing problem, and #91728 is the surviving canonical path that fixes the root cause by making live Copilot /models metadata win over stale static rows for current and future models.

Canonical path: #91728
Source PR: #91875

I am keeping the canonical thread open there so validation and maintainer review stay in one place. Clownfish will preserve this source PR as attribution for the concrete 4.8 symptom report and targeted patch. If this still reproduces after #91728 lands, or if there is a distinct fallback-only case that the canonical PR does not cover, please reply and we can reopen or split it back out.

openclaw-barnacle Bot added extensions: github-copilot size: XS triage: needs-real-behavior-proof Candidate: external PR needs after-fix proof from a real setup. labels Jun 10, 2026

clawsweeper Bot mentioned this pull request Jun 15, 2026

fix(github-copilot): claude-opus-4.8 is native 1M context + thinking (not 128K) #91870

Closed

openclaw-clownfish Bot closed this Jun 15, 2026

openclaw-clownfish Bot added the clownfish Tracked by Clownfish automation label Jun 15, 2026

saju01 mentioned this pull request Jun 17, 2026

[Bug]: Thinking level displayed inconsistently almost everywhere; XHIGH support for models should be enhanced. #59416

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(github-copilot): claude-opus-4.8 context window to 1M with reasoning support#91875

fix(github-copilot): claude-opus-4.8 context window to 1M with reasoning support#91875
yichu10c wants to merge 1 commit into
openclaw:mainfrom
yichu10c:fix/91869-claude-opus-4-8-context-window

yichu10c commented Jun 10, 2026

Uh oh!

clawsweeper Bot commented Jun 10, 2026 •

edited

Loading

Uh oh!

saju01 commented Jun 14, 2026

Uh oh!

openclaw-clownfish Bot commented Jun 15, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

yichu10c commented Jun 10, 2026

Summary

Changes

Why

Test Results

Uh oh!

clawsweeper Bot commented Jun 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

saju01 commented Jun 14, 2026

Uh oh!

openclaw-clownfish Bot commented Jun 15, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

clawsweeper Bot commented Jun 10, 2026 •

edited

Loading