On OpenCode Go, throughput spans 460× — Muse Spark tops the chart but four mid-tier models are the sane defaults

Asked:

“Retrieve opencode offer. For every models in their plan, fetch and compares prices and capabilities. End goal: dress an analysis of the best models to use with opencode go.”

All 25 models on OpenCode Go’s $10/month plan, researched 2026-09-03 from the official Go page: estimated monthly requests (490 to 226,600), token prices per million, included monthly usage ($15, $30 or $60), and privacy terms. Recommendation tiers are our analysis, not OpenCode claims.

What you get — the Go offer

$10

per month, verified current price (opencode.ai)

$12 / 5h

rolling usage cap; $30 per week, $60 per month shared limits

$15–$60

included monthly usage per model — not every model gets the 6× headline value

Blocks

at limits unless you enable fallback to a paid Zen balance

Estimated monthly requests, all 25 models (log scale)

daily-driver candidate high-volume / cheap iteration specialist (code / vision) premium escalation only caution: trains on your prompts general / untiered

Requests are OpenCode’s comparative estimates from “typical Go usage patterns”, not guaranteed quotas. Tiers are our analysis. Hover a bar for prices and privacy.

What the chart cannot say

Muse Spark 1.2 Contributor buys its 226,600 requests by training future Meta models on your prompts — it is not ZDR and is region-limited, so MiMo-V2.5 (150,400 requests, $0.14/$0.28 per M) is the safe volume champion. opencode.ai

Only 13 of 25 models get the full $60 included usage; premium models like Kimi K3 ($3/$15, 490 requests) and Grok 4.6 ($2/$6, 845) get $15 — a fraction of the headline 6× value. opencode.ai

Privacy: most models are “not used for training” with 0-day retention, but Grok 4.6 and GPT 5.6 Luna retain data 30 days, and the page dates DeepSeek’s ZDR agreement only through 2026-08-31 — already past — so confirm before sending sensitive code. opencode.ai

Capability evidence for these brand-new 2026 model versions is thin across independent sources — throughput and price are solid facts, quality is not; benchmark on your own repository. julien.cloud

All 25 models — click a header to sort

Model	Tier	Est. req / mo	Price $/M in / out	Usage $	Retention	Training	Source

Routing playbook

Start MiniMax M3 or Qwen3.7 Plus for everyday coding — $60 usage, ~16,000–21,600 requests/month.

Cheap iteration Qwen3.8 Flash ($0.15/$0.47) or MiMo-V2.5 ($0.14/$0.28) for loops, tests and boilerplate.

Code-heavy work Escalate to Kimi K2.7 Code ($0.95/$4, 6,750 requests, $60 usage) for implementation and refactoring.

Screenshots / images DeepSeek V4 Flash Vision Exp only — images billed as input tokens; experimental, not a text default.

Stubborn problems A/B-test Grok 4.6, Qwen3.8 Max, GLM-5.3 or Kimi K3 first — low quotas make them poor defaults.

Avoid for confidential work Muse Spark 1.2 Contributor: prompts may train Meta models, not ZDR, region-limited.

Method: 25 current OpenCode Go models from the official Go docs page (opencode.ai/docs/go), researched 2026-09-03; capability roundup from julien.cloud. Prices are OpenCode’s billing rates per million tokens, not per-model subscription fees; DeepSeek peak hours are 01:00–04:00 and 06:00–10:00 UTC weekdays. Request counts are OpenCode estimates that depend on prompt length and cache behaviour. Capability, context and modality data for the newly named 2026 versions was inconsistently documented and left blank rather than invented. No current $5 introductory promotion was found; $10/month is the verified offer. Cached-token prices and API endpoint IDs were cut for space.

This report was generated automatically by Keenable SELECT at a user's request, from publicly available web sources linked herein. Keenable does not review, verify, or endorse its contents and makes no representation as to accuracy, completeness, or timeliness; AI-based extraction may contain errors. Nothing in this report is investment, legal, financial, or other professional advice. All trademarks and referenced content remain the property of their respective owners; no affiliation or endorsement is implied. To report an error, rights concern, or request removal: legal@keenable.ai.