↓ Skip to main content

OpenCode Go Models

·3104 words·15 mins

Auto-generated page. This page is automatically built by a scheduled CI pipeline that runs twice daily. The data below is a snapshot of five public sources, cross-referenced programmatically.

Last generated: 2026-11-19 11:09 UTC

Data sources:

This is an unofficial tracker. Not affiliated with or endorsed by OpenCode or Anomaly.

If you spot missing or incorrect information, please reach out so I can fix or improve the page generation.

This page tracks every model available through OpenCode Go, the $10/month subscription for open coding models. The pipeline queries the five sources above, merges their data, and regenerates this page.

Looking for a single model? The model details page lists benchmarks, pricing, request estimates and availability for each of the 42 models.

Quick Stats
#

MetricCount
Documented (listed on Go docs page)32
API metadata endpoint (all models in catalog)31
Models.dev registry42
Undocumented (in API metadata, not on docs page)0
Deprecated9
Total tracked42

The 32 documented models match what a typical Go subscriber sees when running /models in the OpenCode TUI. The API metadata endpoint returns the same 31 models: since November 2026 the docs page no longer publishes a separate subscriber shortlist, so its per-model table and the catalog match one-to-one.

Benchmarks & Value
#

Pricing alone does not tell you which model to use. This section adds independent benchmark data from Artificial Analysis — the same evaluations run for the closed frontier models — and ranks the Go catalogue for the Go plan.

Three numbers drive the ranking:

  • Intelligence Index — Artificial Analysis’ composite across reasoning, coding, knowledge and agentic evaluations, out of 100. Higher is better.
  • Req/5h — requests per 5-hour window on Go. On a flat $10 plan every model costs the same, so this is what separates them.
  • Index × Req/5h — capability per window. This is the sort order below.

Value score = Intelligence Index ÷ blended list price (input:output 1:3). It prices the public list price, not your Go subscription, so treat it as a market signal rather than what you will pay.

Benchmark data exists for 19 of the 42 tracked models, and 16 of those have a composite index. Preview, contributor and free-tier variants usually fall through — see benchmark coverage for the gaps.

Where to start
#

ModelOn Go docsIndexReq/5hIndex × req/5hSpeedTTFTValueBest coding eval
MiMo-V2.6-Flashyes26.730,100804,503414.9 s109.1T-B 4.0 22.7%
DeepSeek V4.1 Flash 1017yes18.543,300801,6191461.2 s38.0T-B 4.0 9.1%
GLM-5.3-Flashyes27.86,320175,720573.3 s67.4T-B 4.0 32.8%
GLM-5.3yes23.16,600152,512652.9 s6.3T-B 4.0 34.8%
Claude Haiku 5.5yes29.93,850115,01622719.0 s74.7T-B 4.0 21.7%

On Go docs: yes marks a model on the official Go docs list – what a typical subscriber sees with /models. — means it is in the API catalog but not documented there: usually usable, but it depends on your subscription, so check /models in the TUI.

Speed (tok/s) and TTFT (seconds to first token) are Artificial Analysis’ measurements on each model’s own vendor API, not on OpenCode Go. They are the right tiebreaker when capability and quota are close: a model that answers four times faster changes how an agent session feels, even when the index is a point lower.

Interactive explorer
#

Filter by benchmark tag, cap the price, and click any column header to sort. First click sorts best-first (cheapest-first for price), second click reverses it. Everything runs client-side — no tracking, no requests.

AA Index
Artificial Analysis composite, 0–100. Higher is better.
Value
Index ÷ blended $/1M: capability per dollar.
T-B 4.0
Terminal-Bench 4.0 — task completion in a real terminal. The closest public proxy for a coding agent.
FC 2.0
FrontierCode 2.0 — hard multi-file software engineering.
$/1M
Blended list price per 1M tokens, input:output 1:3. Not the Go price.
Req/5h
Requests per 5-hour window on the Go plan, from the public Go plan page. ∞ = free, unlimited. Blank = no estimate published for it.
Speed
Output tokens per second, measured by Artificial Analysis against the model’s own (first-party) API — not OpenCode Go’s serving path, so treat it as relative. The same open-weight model runs at very different speeds on different hosts.
TTFT
Seconds until the first output token, same measurement. Low is better; reasoning models can still take much longer to produce a first answer token, which the details page reports separately.
SciCode / GPQA
Scientific Python; graduate-level science reasoning.
ModelIndexValueT-B 4.0FC 2.0$/1MReq/5hSpeedTTFT
Qwen4 MaxAlibabaagentic coding+2↗42.911.448.5%26.5%$3.75463566.3 s
Hy4Tencentagentic coding+2↗34.616.637.4%31.4%$2.082,710952.6 s
Kimi K3.1 ProKimiagentic coding+3↗46.55.867.2%25.5%$8.00147313.6 s
MiniMax M3.1 FlashMiniMaxbest value+2↗29.159.625.8%20.8%$0.493,8801281.3 s
DeepSeek V4.1 Flash 1017DeepSeeklong context↗18.538.09.1%6.1%$0.4943,3001461.2 s
Gemini 4 ArgonGoogleagentic coding+3↗47.66.055.1%25.0%$8.005665815.1 s
Claude Haiku 5.5Anthropicbest value+2↗29.974.721.7%23.2%$0.403,85022719.0 s
MiMo-V2.6-ProXiaomilong context↗29.939.334.8%24.0%$0.763,260563.9 s
MiMo-V2.6-FlashXiaomibest value+2↗26.7109.122.7%21.6%$0.2430,100414.9 s
GPT-6 LunaOpenAIbest value+1↗23.558.68.1%18.4%$0.404,23010016.2 s
Grok 4.7SpaceXAIstrong↗31.16.216.2%21.8%$5.001696814.6 s
Step 5 PreviewStepFunlong context↗28.512.533.3%12.0%$2.274261560.9 s
GLM-5.3-FlashZ AIbest value+2↗27.867.432.8%17.6%$0.416,320573.3 s
GLM-5.3Z AIlong context↗23.16.334.8%12.7%$3.656,600652.9 s
LongCat-2.0LongCatlong context↗——0.0%—$0.9711,400583.3 s
Kimi K2.7 CodeKimi↗——1.0%—$3.24994566.3 s
Qwen3.7 PlusAlibabalong context↗——1.0%—$1.304,310562.8 s
MiniMax-M3MiniMaxlong context↗18.919.42.0%21.9%$0.973,210912.3 s
DeepSeek V4 ProDeepSeeklong context↗23.114.014.1%17.1%$1.651,040872.4 s

All benchmark scores
#

Same columns as the key above, sorted by intelligence index. No JavaScript needed.

ModelAA IndexValueT-Bench 4.0FrontierCode 2.0SciCodeGPQABlended $/1MGo req/5hSpeed (tok/s)TTFTTime/Task
Gemini 4 Argon47.66.055.1%25.0%58.9%—$8.005665815.1 s29.1 min
Kimi K3.1 Pro46.55.867.2%25.5%60.8%—$8.00147313.6 s39.3 min
Qwen4 Max42.911.448.5%26.5%——$3.75463566.3 s31.4 min
Hy434.616.637.4%31.4%——$2.082,710952.6 s13.6 min
Grok 4.731.16.216.2%21.8%54.9%—$5.001696814.6 s10.5 min
MiMo-V2.6-Pro29.939.334.8%24.0%60.9%—$0.763,260563.9 s36.0 min
Claude Haiku 5.529.974.721.7%23.2%48.7%—$0.403,85022719.0 s6.5 min
MiniMax M3.1 Flash29.159.625.8%20.8%53.4%—$0.493,8801281.3 s17.9 min
Step 5 Preview28.512.533.3%12.0%58.9%—$2.274261560.9 s12.1 min
GLM-5.3-Flash27.867.432.8%17.6%51.6%91.2%$0.416,320573.3 s39.4 min
MiMo-V2.6-Flash26.7109.122.7%21.6%51.3%—$0.2430,100414.9 s53.1 min
GPT-6 Luna23.558.68.1%18.4%51.7%—$0.404,23010016.2 s7.3 min
GLM-5.323.16.334.8%12.7%42.0%—$3.656,600652.9 s13.4 min
DeepSeek V4 Pro23.114.014.1%17.1%51.0%92.8%$1.651,040872.4 s18.9 min
MiniMax-M318.919.42.0%21.9%47.1%92.9%$0.973,210912.3 s16.7 min
DeepSeek V4.1 Flash 101718.538.09.1%6.1%38.5%—$0.4943,3001461.2 s7.4 min

Models with individual benchmark results but no composite index yet (Artificial Analysis has not finished scoring them):

ModelT-Bench 4.0FrontierCode 2.0SciCode
LongCat-2.00.0%—36.3%
Kimi K2.7 Code1.0%—47.8%
Qwen3.7 Plus1.0%—46.1%

Benchmark coverage
#

No benchmark data exists for 23 of 42 models. Artificial Analysis only evaluates publicly available endpoints, so preview, contributor and free-tier variants usually fall through the cracks:

Beam Free, Cat Jam Free, Crouton Alpha Free, GLM-5.2, GPT-5.6 Luna, Gemini 4 Argon Contributor, Hy3, Hy4 flash Free, Hy4 preview, Iceman Contributor Free, Kimi K3, LongCat 2.5 Free, LongCat 2.5 Preview Free, MiMo V2.5, MiMo V2.5 Pro, MiniMax M3.1 Flash Free, MiniMax-M2.7, Muse Spark 1.2 Contributor, Muse Spark 1.3 Contributor, Qwen3.8 Flash, Qwen3.8 Max, Space Bunny, Step 5 Preview Free

How to read this
#

  • The index is a composite, not a coding score. A model can rank high overall and still lose on agentic coding — check T-B 4.0 before trusting a top-of-table position.
  • Small score differences do not matter. Treat gaps under two index points as noise.
  • Speed is a separate axis. A model can win on index and quota and still feel sluggish: check Speed and TTFT before committing to it for interactive work. AA measures the vendor API, not the Go plan, so compare models against each other rather than trusting the absolute number.
  • A high value score is not a recommendation. It means cheap capability, not the right capability. The cheapest model that fails your task is the most expensive one you can pick.
  • Vendor-reported numbers and independent ones share a column. Where a score is marked as vendor-reported, treat it as an upper bound.

Scores are refreshed when the tracker runs and are sourced from Artificial Analysis; the full methodology explains how the index is built.

Changelog
#

Changes detected since previous snapshot (2026-11-16).

Removed from API
#

DateModel
2026-11-19GLM-5.2 removed from OpenCode Go API

Marked as Deprecated
#

DateModel
2026-11-19GLM-5.2 marked as deprecated

Removed Models
#

No longer available from any source. Dates come from the tracked history.

ModelRemoved
DeepSeek V4 Flash2026-11-10
GLM-52026-11-10
GLM-5.12026-11-10
Hy3 Preview2026-11-10
Kimi K2.52026-11-10
Kimi K2.62026-11-10
MiMo V2 Omni2026-11-10
MiMo V2 Pro2026-11-10
MiniMax M2.52026-11-10
Qwen3.5 Plus2026-11-10
Qwen3.6 Plus2026-11-10
Qwen3.7 Max2026-11-10

Tracked History
#

Every addition and removal recorded since tracking began (5 dates), newest first.

DateChangeModels
2026-11-16AddedHy4 flash Free, Qwen4 Max
2026-11-10AddedBeam Free, Cat Jam Free, Claude Haiku 5.5, Crouton Alpha Free, DeepSeek V4.1 Flash 1017, GLM-5.2, +28 more
2026-11-10RemovedDeepSeek V4 Flash, GLM-5, GLM-5.1, Hy3 Preview, Kimi K2.5, Kimi K2.6, +6 more
2026-06-04AddedQwen3.7 Plus
2026-06-02AddedDeepSeek V4 Flash, DeepSeek V4 Pro, GLM-5, GLM-5.1, Hy3 Preview, Kimi K2.5, +11 more

All Models
#

ModelDocumentedStatusReleaseContextInput $/1MOutput $/1MCache ReadReq/5hSources
Qwen4 Maxyesactive2026-11-161M$1.50$4.50$0.07463API Models.dev Docs
Hy4 flash Freeyesactive2026-11-161M$0$0$0∞API Models.dev Docs
Hy4yesactive2026-11-041M$0.83$2.50$0.042,710API Models.dev Docs
Crouton Alpha Freeremoveddeprecated2026-11-031M$0$0$0—Models.dev
LongCat 2.5 Freeyesactive2026-11-021M$0$0$0∞API Models.dev Docs
Iceman Contributor Freeyesactive2026-10-24262K$0$0$0∞API Models.dev Docs
Kimi K3.1 Proyesactive2026-10-231M$2.00$10.00$0.20147API Models.dev Docs
MiniMax M3.1 Flash Freeremoveddeprecated2026-10-191M$0$0$0—Models.dev
MiniMax M3.1 Flashyesactive2026-10-191M$0.15$0.60$0.013,880API Models.dev Docs
DeepSeek V4.1 Flash 1017yesactive2026-10-171M$0.15$0.60$0.00343,300API Models.dev Docs
Gemini 4 Argon Contributorremoveddeprecated2026-10-131M$4.00$20.00$0.20—Models.dev
Gemini 4 Argonyesactive2026-10-131M$2.00$10.00$0.10566API Models.dev Docs
Cat Jam Freeremoveddeprecated2026-10-121M$0$0$0—Models.dev
Claude Haiku 5.5yesactive2026-10-071M$0.10$0.50$0.013,850API Models.dev Docs
Beam Freeyesactive2026-10-051M$0$0$0∞API Models.dev Docs
LongCat 2.5 Preview Freeremoveddeprecated2026-09-251M$0$0$0—Models.dev
Space Bunnyremoveddeprecated2026-09-231M$0.15$0.60$0.03—Models.dev
MiMo-V2.6-Proyesactive2026-09-221M$0.43$0.87$0.00363,260API Models.dev Docs
MiMo-V2.6-Flashyesactive2026-09-221M$0.14$0.28$0.002830,100API Models.dev Docs
GPT-6 Lunayesactive2026-09-221M$0.10$0.50$0.014,230API Models.dev Docs
Grok 4.7yesactive2026-09-21500K$2.00$6.00$0.50169API Models.dev Docs
Step 5 Preview Freeremoveddeprecated2026-09-161M$0$0$0—Models.dev
Step 5 Previewyesactive2026-09-161M$1.00$2.70$0.10426API Models.dev Docs
Muse Spark 1.3 Contributoryesactive2026-09-021M$0.10$0.20$0.00245,300API Models.dev Docs
Hy4 previewyesactive2026-08-281M$0.83$2.50$0.041,350API Models.dev Docs
Qwen3.8 Flashyesactive2026-08-261M$0.15$0.47$0.025,400API Models.dev Docs
GLM-5.3-Flashyesactive2026-08-261M$0.15$0.50$0.036,320API Models.dev Docs
GLM-5.3yesactive2026-08-141M$1.40$4.40$0.266,600API Models.dev Docs
Muse Spark 1.2 Contributoryesactive2026-08-051M$0.10$0.20$0.00245,300API Models.dev Docs
Qwen3.8 Maxyesactive2026-08-031M$2.00$6.00$0.25162API Models.dev Docs
Kimi K3yesactive2026-07-161M$3.00$15.00$0.3098API Models.dev Docs
GPT-5.6 Lunayesactive2026-07-091M$0.20$1.20$0.022,050API Models.dev Docs
Hy3yesactive2026-07-06256K$0.14$0.58$0.044,300API Models.dev Docs
LongCat-2.0yesactive2026-06-301M$0.30$1.20$0.00611,400API Models.dev Docs
GLM-5.2yesdeprecated2026-06-131M$1.40$4.40$0.26792Models.dev Docs
Kimi K2.7 Codeyesactive2026-06-12262K$0.95$4.00$0.19994API Models.dev Docs
Qwen3.7 Plusyesactive2026-06-021M$0.40$1.60$0.044,310API Models.dev Docs
MiniMax-M3yesactive2026-06-011M$0.30$1.20$0.063,210API Models.dev Docs
DeepSeek V4 Proyesactive2026-04-241M$0.66$1.98$0.021,040API Models.dev Docs
MiMo V2.5 Proremovedactive2026-04-221M$0.43$0.87$0.0036—Models.dev
MiMo V2.5removedactive2026-04-221M$0.14$0.28$0.0028—Models.dev
MiniMax-M2.7removeddeprecated2026-03-18205K$0.30$1.20$0.06—Models.dev

Model details
#

Benchmarks, pricing, request estimates and source availability for every model are on the model details page.

Undocumented Models
#

None this run: every model in the API metadata endpoint also appears in the model table on the official Go documentation page. The page previously published a shorter subscriber list; since November 2026 the docs table covers the whole catalog, so this split no longer separates restricted models.

Deprecated Models
#

These models are marked as deprecated in Models.dev. They may still be available through the API but are likely to be removed.

ModelRelease DateReplacement
Crouton Alpha Free2026-11-03—
MiniMax M3.1 Flash Free2026-10-19—
Gemini 4 Argon Contributor2026-10-13—
Cat Jam Free2026-10-12—
LongCat 2.5 Preview Free2026-09-25—
Space Bunny2026-09-23—
Step 5 Preview Free2026-09-16—
GLM-5.22026-06-13—
MiniMax-M2.72026-03-18—

Source Availability Matrix
#

ModelAPIModels.devDocsStatus
Qwen4 Maxyesyesyesactive
Hy4 flash Freeyesyesyesactive
Hy4yesyesyesactive
Crouton Alpha Free—yes—deprecated
LongCat 2.5 Freeyesyesyesactive
Iceman Contributor Freeyesyesyesactive
Kimi K3.1 Proyesyesyesactive
MiniMax M3.1 Flash Free—yes—deprecated
MiniMax M3.1 Flashyesyesyesactive
DeepSeek V4.1 Flash 1017yesyesyesactive
Gemini 4 Argon Contributor—yes—deprecated
Gemini 4 Argonyesyesyesactive
Cat Jam Free—yes—deprecated
Claude Haiku 5.5yesyesyesactive
Beam Freeyesyesyesactive
LongCat 2.5 Preview Free—yes—deprecated
Space Bunny—yes—deprecated
MiMo-V2.6-Proyesyesyesactive
MiMo-V2.6-Flashyesyesyesactive
GPT-6 Lunayesyesyesactive
Grok 4.7yesyesyesactive
Step 5 Preview Free—yes—deprecated
Step 5 Previewyesyesyesactive
Muse Spark 1.3 Contributoryesyesyesactive
Hy4 previewyesyesyesactive
Qwen3.8 Flashyesyesyesactive
GLM-5.3-Flashyesyesyesactive
GLM-5.3yesyesyesactive
Muse Spark 1.2 Contributoryesyesyesactive
Qwen3.8 Maxyesyesyesactive
Kimi K3yesyesyesactive
GPT-5.6 Lunayesyesyesactive
Hy3yesyesyesactive
LongCat-2.0yesyesyesactive
GLM-5.2—yesyesdeprecated
Kimi K2.7 Codeyesyesyesactive
Qwen3.7 Plusyesyesyesactive
MiniMax-M3yesyesyesactive
DeepSeek V4 Proyesyesyesactive
MiMo V2.5 Pro—yes—active
MiMo V2.5—yes—active
MiniMax-M2.7—yes—deprecated

About This Page
#

This page is automatically generated by a scheduled CI pipeline that queries five publicly accessible sources:

  • OpenCode Go API – the unauthenticated metadata endpoint listing all models in the Go catalog
  • Models.dev – provider/model metadata including costs, release dates, and deprecation status
  • OpenCode Go Documentation – the official docs page, source of the documented model list
  • OpenCode Go console – the public Go plan page, source of the per-model request estimates
  • Artificial Analysis – independent benchmark scores and public API pricing used for the value ranking

What this tracker can and cannot tell you
#

What it verifies: which models exist in each public data source, their release dates, pricing, deprecation status, and when models are added to or removed from those sources.

What it cannot verify: which models your specific subscription entitles you to. The API metadata endpoint is unauthenticated and lists models that may not be available to all subscribers. To see which models you can actually use, run /models in the OpenCode TUI – that command authenticates with your API key.

The “Documented” column above reflects which models appear in the model table on the official docs page – those typically match what a standard Go subscription includes, but this is not independently verified by this tracker. Since November 2026 that table covers the whole catalog, so it no longer separates restricted models from standard ones.

This is an unofficial tracker. It is not affiliated with or endorsed by OpenCode or Anomaly.

See Also
#