Feature Request
Description
Google announced Gemini 3.5 Flash today (May 19, 2026) at Google I/O. This is their latest frontier model combining strong intelligence with exceptional speed, and it is now generally available via the Gemini API in Google AI Studio.
I'd like to request first-class support for Gemini 3.5 Flash (and the upcoming 3.5 Pro) as a model option in OpenCode.
Why Gemini 3.5 Flash?
- Frontier-level coding performance: Outperforms Gemini 3.1 Pro on Terminal-Bench 2.1 (76.2%), and leads in agentic benchmarks like GDPval-AA (1656 Elo) and MCP Atlas (83.6%).
- 4x faster output tokens/sec compared to other frontier models — ideal for an interactive coding agent like OpenCode.
- Strong agentic capabilities: Designed for complex, long-horizon agentic tasks — planning, building, and iterating across multi-step workflows.
- Multimodal understanding: Leads in CharXiv Reasoning (84.2%), useful for image/diagram understanding in coding contexts.
- Cost-efficient: Often less than half the cost of other frontier models for comparable quality.
- Already available via API: Generally available through Google AI Studio and the Gemini API today.
Expected model IDs
- \gemini-3.5-flash\ (available now)
- \gemini-3.5-pro\ (expected next month)
Reference
Additional Context
Gemini 3.5 Flash is positioned in the top-right quadrant of the Artificial Analysis index (frontier intelligence + exceptional speed), making it an excellent fit for OpenCode's interactive coding agent workflow where both quality and latency matter.
Feature Request
Description
Google announced Gemini 3.5 Flash today (May 19, 2026) at Google I/O. This is their latest frontier model combining strong intelligence with exceptional speed, and it is now generally available via the Gemini API in Google AI Studio.
I'd like to request first-class support for Gemini 3.5 Flash (and the upcoming 3.5 Pro) as a model option in OpenCode.
Why Gemini 3.5 Flash?
Expected model IDs
Reference
Additional Context
Gemini 3.5 Flash is positioned in the top-right quadrant of the Artificial Analysis index (frontier intelligence + exceptional speed), making it an excellent fit for OpenCode's interactive coding agent workflow where both quality and latency matter.