OpenCode coding plans

Community

Value snapshot · Grok 4.6 · typical shape · req / $1

One dollar on OpenCode buys 85 Grok 4.6 requests at its best plan — strongest value we track for this flagship. Exact-priced flagship only — no proxy claims in the headline.

Best for

27 events in the last 90 days — 7 worse, 16 better.

• Policy20 Oct 2026

  OpenCode’s referral programme has ended, so the site drops the referral metadata it was still shipping. The provider states it on its own Go page: any URL carrying a `ref` parameter renders the warning banner "The referral program has ended. Referral links no longer earn credit for you or the person who shared them." (i18n key `go.referral.ended`, label `Warning`, in the page’s own bundle and rendered whenever `searchParams.ref` is set). The bare `/go` page carries no referral copy at all, the docs (`/docs/go`, `/docs/zen`) and the console bundle contain zero occurrences of "referral", and there is no referral page in the sitemap — there is no surviving public programme to re-apply to. No plan price ($10 Go / $40 Go Plus), cap, roster, `modelsUrl` or `lastChecked` value moved; `codingTool.lock` stays `open`.

  Verified 2026-10-20 · source

• Ceiling14 Oct 2026

  The published per-model monthly limit for DeepSeek V4.1 Flash moves $60 → $80 on Go and $120 → $160 on Go Plus, at the same off-peak token rates ($0.15 input / $0.60 output / $0.003 cached read per 1M, peak exactly double). The Go record’s headline follows its own top-of-range convention, so the plan-level figure moves $60/mo → $80/mo with the published fractions giving $40/week and $16 per 5 hours; Go Plus keeps its $240/mo headline, which is still the top of its $60–$240 range. The page’s request estimates for the model move with the limit ($173,330 requests per month on Go, $346,670 on Go Plus at the published shape). Plan prices ($10 Go / $40 Go Plus), every per-token rate and every other per-model limit are unchanged.

  Verified 2026-10-14 · source

• Models14 Oct 2026

  Mixed roster change in one read. The page no longer carries DeepSeek V4 Flash or DeepSeek V4 Flash Vision Exp anywhere — not in the advertised model list, the usage-limit table, the request-estimate table or the endpoint table — so both leave the two tiers’ rosters and their per-model rules; the names stay in the catalog because other tracked plans still serve them. In the same read the page added Gemini 4 Argon Contributor (endpoint id gemini-4-argon-contributor) to both tiers at $4.00 input / $20.00 output / $0.20 cached read per 1M with a $30 monthly limit on Go and $120 on Go Plus, doubled for a limited time under a “2x” label — the plan notes describe the doubling and the standing figure is the one the calculator prices. The page’s own data policy for the row reads model training in use and retention not zero-data-retention. Direction is neutral because the read both adds and removes model access: a frontier Gemini arrives while two older DeepSeek Flash variants leave.

  Verified 2026-10-14 · source

• Ceiling09 Oct 2026

  Site disclosure: opencode.ai/docs/go publishes a per-model “Monthly limit” column (Go $15–$60, Go Plus $60–$240) and both records carried only its DeepSeek and Grok rows, so the notes now carry the published window fractions and name the models that cap below the plan headline — GLM-5.3, Kimi K3, Qwen3.8 Max, the MiMo V2.6/V2.5 Pro rows, GPT-6 and GPT-5.6 Luna and Claude Haiku 5.5 at $15 on Go ($60 on Go Plus), where a few flagship requests exhaust the allowance. No price, per-token rate, plan ceiling, roster or per-model cap moved: every figure named was already on the page and the plan-level headline limits are unchanged.

  Verified 2026-10-09 · source

• Models08 Oct 2026

  OpenCode Go’s docs page prices a Claude Haiku 5.5 row that the plan’s roster did not carry, and OpenCode’s own router catalogue serves it, so the model is a live entitlement rather than a page-only listing. Added to both tiers. The ≤ 100K-token band reads $0.10 input / $0.50 output / $0.01 cached read / $0.125 cached write per 1M at a $15 monthly limit, with the > 100K band at $0.50 / $2.50 / $0.05 / $0.625 and a $60 limit on Go Plus; the page’s request shape is 1,100 input / 55,000 cached / 240 output tokens per request, and its endpoint row routes to OpenCode’s Anthropic Messages endpoint with request retention of 30 days under Anthropic’s data policies (training: not used). The $10 / $40 plan prices, every existing rate and every other per-model limit are unchanged.

  Verified 2026-10-08 · source

Head-to-head

OpenCode compared side by side

Two plans side by side — shared models, BYOK reachability and value per $1, generated from the same verified data as the calculator.

• OpenCode GovsCommandCode
• OpenCode GovsQwenCloud
• OpenCode GovsOzore
• Nous PortalvsOpenCode Go
• ClinevsOpenCode Go
• NanoGPTvsOpenCode Go

Flagship rank · Grok 4.6: #28 (xHigh config) on the Agent Arena agent leaderboard — 0.89% net improvement (board fetched 2026-10-19).

Common questions