OpenCode has launched OpenCode Go, a $10/month subscription tier that bundles curated open coding models behind a single API key (endpoint opencode.ai/zen/go/v1). The page positions Go as an optional provider inside the OpenCode TUI, activated via /connect, with model selection listed through /models, and it is framed The Go subscription carries dollar-denominated usage caps rather than request quotas: $12 per 5-hour window, $30 per week, and $60 per month, with the effective request count varying by model price. The current model roster spans Grok 4.6, the GLM-5.x family, GPT 5.6 Luna, Kimi K2.7/K2.6 Code, LongCat-2.0, MiMo-V2.5 an

Using any LLM provider in OpenCode.

The official OpenCode changelog ships a provider-layer change that directly affects OpenCode Go users: web search is now enabled when going through the OpenCode Go provider, restoring parity with other providers that already expose search-augmented completions. The same changelog entry also lists broader provider fixes Additional Go-relevant provider adjustments include retry logic that resumes provider responses ending with finish_reason: network_error (covering more network-error variants) and retries on xAI capacity/temporary-unavailability stream errors, which improve reliability for OpenCode Go customers hitting those upstreams.

According to an OpenCode engineer quoted by KuCoin/InfoQ, DeepSeek's upcoming price increase is likely a traffic management move rather than a sign of operating losses, since current demand is already straining server capacity. The engineer also noted that OpenCode can in principle match DeepSeek's pricing, but OpenCod The gap is attributed to cache efficiency: DeepSeek's official API can reach near-100% cache hit rates within 12 hours, while OpenCode Go accumulates more unhit caches over extended sessions, amplifying token spend. The advantage comes from DeepSeek's prefix-based disk persistence mechanisms combined with V4's Token-wi