OpenCode v1.14.25 Adds Roslyn LSP Support for Razor and Fixes GPT-5.5 Context Limits
OpenCode shipped v1.14.25 with Roslyn LSP support for Razor, .cshtml, and C# script files. The release fixes GPT-5.5 with OpenAI OAuth to use correct context limits and adds request details to LSP permission prompts.
OpenCode released v1.14.25 with support for Roslyn Language Server Protocol (LSP) across Razor, .cshtml, and C# script files.
Core Changes
The release addresses a context compaction issue where GPT-5.5 with OpenAI OAuth was not using correct context limits. The fix ensures the model operates within its specified token boundaries.
LSP permission prompts now include granular request details: the specific operation being performed, target file path, and cursor position. This gives developers visibility into what code assistance tools are accessing before granting permission.
The permission configuration system now preserves rule order and exposes full IntelliSense support for tool permission keys, improving the developer experience when configuring access controls.
Shell commands maintain the correct working directory after login shell startup files execute, fixing a bug where the working directory would change unexpectedly during shell initialization.
Community Contributions
Two community members contributed to this release:
- @altendky clarified git amend conditions to require verifying commits have landed (PR #19937)
- @b0o added Prettier to devDependencies for the build system (PR #23255)
What This Means
The Roslyn LSP integration expands OpenCode's C# development capabilities beyond standard .cs files into web development with Razor templates and scripting scenarios. The GPT-5.5 context fix is critical for developers using OpenAI's OAuth flow, as incorrect context limits can cause the model to prematurely compress or drop relevant code context. Enhanced permission prompts with operation details give developers better control over what their AI coding assistants can access, addressing security and privacy concerns in enterprise environments.
Related Articles
OpenAI Python SDK v3.8.0 Reveals Reference to Unannounced 'gpt-6-astra' Model
The openai-python SDK v3.8.0 release notes, dated September 3, 2026, list a feature addition for 'gpt-6-astra' — a model name not previously confirmed by OpenAI. No official announcement, pricing, or specifications have been released.
Simon Willison's Pelican Benchmark Shows GPT-6 Astra Outperforming GPT-5.6 Sol at Every Reasoning Level
Developer Simon Willison ran his signature 'pelican riding a bicycle' SVG test on newly-accessed GPT-6 Astra across five reasoning levels, comparing results against GPT-5.6 Sol, Terra, and Luna. Even Astra's lowest reasoning setting reportedly beat every Sol output, though Astra costs roughly twice as much per token.
OpenAI Lists GPT-6 Astra Pro on OpenRouter: Same Model, Higher-Compute Reasoning Mode
GPT-6 Astra Pro, now listed on OpenRouter, is the existing GPT-6 Astra model configured to run with reasoning.mode set to 'pro' for higher-quality output on complex tasks. It carries a 1M-token context window and tiered pricing from $5/$25 to $20/$100 per million input/output tokens depending on the serving tier.
OpenAI's GPT-6 Astra Cuts Hallucinations, But Indirect Prompt Injection Attacks Still Succeed 8.5% of the Time
OpenAI's new GPT-6 Astra model shows major improvements in hallucination rates and jailbreak resistance over predecessor GPT-5.6 Sol, according to OpenAI's system card. However, indirect prompt injection attacks hidden in documents still succeed 8.5% of the time in external testing by Gray Swan, down from 27% but still above rival Claude Opus 5's 4.8% rate.
Comments
Loading...