OpenCode

Added the Astra system prompt for GPT-6 models.

• Preserved Bedrock DeepSeek model IDs, including ARN-based IDs, so they resolve correctly. (@YeEmrick)
• Updated the Azure provider SDK to pick up compatibility fixes.
• Updated the OpenAI provider SDK to pick up compatibility fixes.
• Added reasoning effort variants for supported GitLab GPT and Claude models. (@far-ouq)
• **Thank you to 3 community contributors:**
• @heimoshuiyu:
• fix(console): openai usage normalization and tier threshold config (#47342)
• @far-ouq:
• fix(opencode): add GitLab reasoning variants (#47306)
• @YeEmrick:
• +1 more change

Allow Codex OAuth model filtering to recognize integer GPT versions like `gpt-6`

• Fixes issue of gpt-6-astra not showing up for openai subscription users
• **Thank you to 2 community contributors:**
• @Peter267:
• docs(zh): fix bold rendering by adding spaces around asterisks (#46231)
• @heimoshuiyu:
• feat(console): add quota reset support action (#47313)

Send the session ID as GitHub Copilot's interaction header to improve request tracking across a session.

• Use the desktop client ID during OpenCode account device authentication.
• Increase the open-in app icon size for better visibility.
• **Thank you to 1 community contributor:**
• @MVS-source:
• docs: add Eden AI to the providers list (#43386)

Default provider header timeouts to five minutes so slow model startups fail less often.

• Default streamed chunk timeouts to five minutes, with `false` supported to disable them.
• Let Anthropic `thinking.blockBinding` opt out via config when you need to keep the provider default. (@dkindlund)
• Limit Anthropic thinking block binding to Claude 5.1+ models so older deployments do not reject requests.
• Avoid unhandled errors when canceling timed-out SSE reads. (@AlexanderWillner)
• **Thank you to 2 community contributors:**
• @AlexanderWillner:
• +3 more changes

Claude 5 sessions now tolerate stale thinking blocks instead of failing after prompt or tool changes.

• Bedrock GPT-5.6 models now accept `none` reasoning effort.
• Bedrock reasoning and replay handling is more reliable. (@pengzh1)
• Tool call timing now stays accurate when tools update their metadata while still running. (@bartlettroscoe)
• `apply_patch` no longer emits an empty move path in permission metadata. (@altendky)
• Azure CLI sign-in now asks for the resource name directly instead of querying Azure management APIs.
• +8 more changes

Fixed Azure authentication so Azure CLI sign-in works without requiring Bun.

Bedrock reasoning responses no longer get cached into unreplayable empty messages.

• Azure providers can now sign in with Microsoft Entra ID through the Azure CLI instead of requiring an API key.
• V1 now reads supported V2 config fields so newer config files keep working in more mixed setups.
• Archived sessions disappear from the Home list immediately. (@NathanTCode)
• **Thank you to 3 community contributors:**
• @NathanTCode:
• fix(app): drop archived sessions from home list right away (#44905)
• @OpeOginni:
• fix(ui): restore focus in stacked dialogs (#44928)
• @dondetir:
• +1 more change

Fixed Cloudflare AI Gateway routing for third-party providers so non-Workers models work through the gateway's REST API. (@superhighfives)

• Fixed Anthropic models through Cloudflare AI Gateway by converting dotted model IDs like `claude-haiku-4.5` to the dashed slug Anthropic expects. (@superhighfives)
• Fixed parent session IDs being sent in request headers for session-aware providers.
• Fixed GitHub auth for immutable OIDC subject tokens.
• **Thank you to 1 community contributor:**
• @superhighfives:
• fix(provider): send Anthropic's dashed native slug through the AI Gateway (#44281)
• +1 more change

Removed outdated OpenCode Go first-month discount messaging and pricing.

• Fixed OpenCode device login links when servers return relative verification URLs or use a base path.
• Fixed `textVerbosity` being sent to OpenAI-compatible providers that do not support it. (@joelstucki-taulia)
• Updated the Amazon Bedrock provider for compatibility fixes.
• Keep model provider headers visible while scrolling the model picker.
• **Thank you to 1 community contributor:**
• @joelstucki-taulia:
• fix(provider): guard textVerbosity injection for @ai-sdk/openai-compatible providers (#43915)

Continue responses when a model reports an unknown finish reason instead of stopping early

• Route Vertex AI `eu` and `us` multi-region Gemini requests through REP endpoints
• Keep file search results visible while the next search is loading
• Register the archive session command in both desktop layouts (@NathanTCode)
• **Thank you to 1 community contributor:**
• @NathanTCode:
• fix(app): register archive session command in both layouts (#41741)

Surface failed subagent tool calls with a resumable `task_id`.

• Retry provider responses that end with `finish_reason: network_error`.
• Retry more network error variants, including `network-error` and `network_error`.
• Surface resumable subagent failures instead of returning an empty result.
• Preserve Cerebras `max_completion_tokens` without applying an extra output cap.
• Answer permission requests triggered by subagents during `opencode run`.
• Retry xAI capacity and temporary unavailability stream errors.

Added native OpenAI and Anthropic passthroughs for Cloudflare AI Gateway models.

• Matched Codex rate limits more closely to ChatGPT subscription limits. (@GameOn223)
• Removed built-in Qwen sampling defaults that could send unsupported settings.
• Properly show authenticated providers in `/connect`. (@OpeOginni)
• Ignore malformed model pricing instead of breaking usage cost calculation.
• Fall back when OpenAI websocket requests exceed message size limits.
• Forward ChatGPT workspace compute residency to Codex requests.
• Updated the default Console URL used by account connection flows.
• +12 more changes

Select the Kimi system prompt correctly for official Moonshot and Kimi providers

• Fix xhigh reasoning effort for xai models

Made session compaction keep complete recent turns and produce clearer summaries for smaller models.

• Added MERGE Gateway reasoning variants so those model options work correctly. (@MatthewFeroz)
• Capped automatic session retries and added jitter to reduce repeated retry storms.
• Enabled PDF attachments for GitHub Copilot models that advertise PDF vision support. (@stevenao)
• Applied the correct sampling defaults to DeepSeek V4 Flash on supported providers.
• Routed all Muse family models to the correct Meta system prompt. (@mreso)
• +16 more changes

Ignore unknown top-level config fields instead of failing config parsing

• Register projects opened from Home so they are available to the rest of the app
• Open the project menu with a right-click in Home
• Fall back to listing and matching local directories when the project picker server does not support search
• Use 词元 instead of 令牌 for token-related labels in Simplified Chinese (@Speechlessmanbilibili)
• Keep the macOS app running after the last window closes and reopen a window when the app is activated
• **Thank you to 1 community contributor:**
• @Speechlessmanbilibili:
• +1 more change