# ZenPick > ZenPick compares every OpenCode Go model side-by-side with live benchmarks, pricing, quota-burn estimates, and algorithmic task-fit scores — so developers pick the right model for the job before spending their subscription quota. ZenPick is a free, read-only comparison tool for the [OpenCode Go](https://opencode.ai/go) model catalog (the $10/month tier). It aggregates public data from modelgrep, LLM Stats, and the OpenCode Go docs, then ranks each model per task. The catalog is interactive on the site (JavaScript-rendered); the table below is the same catalog in plain markdown. - Models: 28 - View: ranked for Cheapest, best first - Data as of: 2026-10-23T12:31:40.445Z - Generated: 2026-10-23T12:31:40.450Z ## How to use this file This one file answers "which OpenCode Go model should I use". It is regenerated on every request from the same cache the site reads (at most ~6 hours stale), so a single fetch is current. No browser and no API key required. **1. Read the answers before the data.** `Quick answers` below is this site's own ranking per leaderboard: `coding`, `design`, `smartest`, `reasoning`, `vision`, `open-source`, `long-context`, and `cheapest`. Use those lines instead of recomputing: tied leaders, fit weighting, and models that have no data for a metric are already handled. **2. Search by filtering what you already fetched.** The catalog is small enough to hold in context: grep or sort the table by the column the question is about. `Input $/M` for cost, `Context` for window size, `Coding` or `Agentic` for capability, `Burn` for quota drain, `Replaces` for which closed model it substitutes. On the site, `?q=` searches model name, provider, tags, and capability words. Both parameters work on this file too (`?need=cheapest`, `?q=vision`), which sorts or filters the table before you read it. **3. Quote the `ID`, not the name.** IDs like `deepseek-v4.1-flash` are the stable keys used by the site, its compare URLs, and this file; names are display text and change. **4. Know what is not in the table.** Input/output modalities (image, video, audio, file), tool calling, structured output, and reasoning flags are not columns here. For those, open the model's LLM Stats link in the `Source` column, or fetch the `raw` slice (see **Query URLs**), which returns the full model objects including `capabilities`. **5. When this file is not enough:** - **Methodology** (how scores, burn bands, and replacement hints are computed): `https://zp.svelte-apps.me/about`. - **The upstream docs**, whose numbers this file summarizes: [OpenCode Go docs](https://opencode.ai/v2/docs/console/go) is the source of truth for the official model list, prices, and usage limits, so it wins on anything contractual; [modelgrep](https://modelgrep.com) holds OpenRouter pricing plus the Artificial Analysis and Design Arena benchmarks; [LLM Stats](https://llm-stats.com) holds per-model metadata, releases, and capabilities. - **The site itself**: the homepage renders in the browser, so fetching its HTML returns a shell rather than the catalog. A client that runs JavaScript can open a leaderboard or comparison URL directly; **Query URLs** below lists them. **6. Cite the freshness.** When quoting a number, link `https://zp.svelte-apps.me/llms.txt` and mention the `Data as of` timestamp above, so a reader can judge how current the answer is. ## Quick answers One line per leaderboard: - Qwen3.8 Max and Kimi K3 are tied as the best opencode models for coding — 76.2 Coding Index. - Kimi K3 is the best opencode model for design & frontend work — 1389 Design Arena Elo. Gemini 4 Argon Contributor (1387) and DeepSeek V4.1 Flash (1357) round out the top three. 13 more models have no design arena elo data and are ranked last. - Muse Spark 1.2 Contributor is the best opencode model for general intelligence — 80.0 Intelligence Index. Qwen3.8 Flash (70.0) and Hy4 preview (68.0) round out the top three. 2 more models have no intelligence index data and are ranked last. - Muse Spark 1.2 Contributor is the best opencode model for step-by-step reasoning — 80.0 Intelligence Index. Qwen3.8 Flash (70.0) and Hy4 preview (68.0) round out the top three. 1 more model has no intelligence index data and is ranked last. - Muse Spark 1.2 Contributor is the best opencode model for vision & document understanding — 80.0 Intelligence Index. Qwen3.8 Flash (70.0) and Gemini 4 Argon Contributor (52.6) round out the top three. 1 more model has no intelligence index data and is ranked last. - Hy4 preview is the best opencode model for self-hosted deployment — 68.0 Intelligence Index. MiMo-V2.6-Pro (46.3) and GLM-5.3 (44.8) round out the top three. 1 more model has no intelligence index data and is ranked last. - MiMo-V2.6-Pro, GPT-6 Luna, MiMo-V2.6-Flash, and 17 more are tied for the largest context window at 1.05M tokens. - Beam Free is the cheapest opencode model at $0.00 per million input tokens. ## Models | Rank | Model | Input Price / 1M | ID | Provider | Context | Output cap | Reasoning efforts | Knowledge cutoff | Status | Input $/M | Output $/M | Burn | Req/mo | Coding | Agentic | Replaces | Source | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | 1 | Beam Free | $0.00 | beam-free | Unknown | 128K | 131K | low, medium, high, xhigh, max | 2026-06-30 | active | Free | Free | Free | Unlimited | 0 | 0 | — | [llm-stats](https://llm-stats.com/models/beam-free) 2 | Muse Spark 1.2 Contributor | $0.10 | muse-spark-1.2-contributor | meta | 1.0M | 131K | minimal, low, medium, high, xhigh | — | active | $0.10 | $0.20 | Excellent | 226,600 | 70 | 70 | — | [llm-stats](https://llm-stats.com/models/muse-spark-1.2) 2 | Claude Haiku 5.5 | $0.10 | claude-haiku-5-5 | anthropic | 1M | 128K | low, medium, high, xhigh, max | 2026-06 | active | $0.10 | $0.50 | Moderate | 19,230 | 67 | 67 | — | [llm-stats](https://llm-stats.com/models/claude-haiku-5-5) 2 | GPT-6 Luna | $0.10 | gpt-6-luna | openai | 1.1M | 128K | none, low, medium, high, xhigh, max | 2026-05-18 | active | $0.10 | $0.50 | Good | 21,130 | 52 | 54 | — | [llm-stats](https://llm-stats.com/models/gpt-6-luna) 2 | Muse Spark 1.3 Contributor | $0.10 | muse-spark-1.3-contributor | meta | 1.0M | 131K | minimal, low, medium, high, xhigh | — | active | $0.10 | $0.20 | Excellent | 226,600 | 70 | 70 | — | [llm-stats](https://llm-stats.com/models/muse-spark-1.3) 6 | MiMo-V2.6-Flash | $0.14 | mimo-v2.6-flash | xiaomi | 1.1M | 131K | — | — | active | $0.14 | $0.28 | Excellent | 150,400 | 41 | 45 | GPT-6 Luna | [llm-stats](https://llm-stats.com/models/mimo-v2.6-flash) 6 | Hy3 | $0.14 | hy3 | tencent | 262K | 128K | none, low, high | — | active | $0.14 | $0.58 | Good | 21,500 | 50 | 30 | Claude Haiku 5.5, Grok 4.3 | [llm-stats](https://llm-stats.com/models/hy3) 6 | MiMo V2.5 | $0.14 | mimo-v2.5 | xiaomi | 1.1M | 128K | — | 2024-12 | active | $0.14 | $0.28 | Excellent | 150,400 | 60 | 60 | Claude Haiku 5.5, Gemini 3.5 Flash-Lite | [llm-stats](https://llm-stats.com/models/mimo-v2.5) 9 | Qwen3.8 Flash | $0.15 | qwen3.8-flash | qwen | 1M | 131K | low, medium, xhigh | — | active | $0.15 | $0.47 | Good | 27,000 | 60 | 60 | — | [llm-stats](https://llm-stats.com/models/qwen3.8-flash) 9 | MiniMax M3.1 Flash | $0.15 | minimax-m3.1-flash | minimax | 1.0M | 512K | low, medium, high, xhigh, max | — | active | $0.15 | $0.60 | Moderate | 19,380 | 51 | 55 | — | [llm-stats](https://llm-stats.com/models/minimax-m3.1-flash) 9 | GLM-5.3-Flash | $0.15 | glm-5.3-flash | z-ai | 1.0M | 131K | low, high, max | — | active | $0.15 | $0.50 | Good | 31,580 | 56 | 60 | Claude Sonnet 5, Claude Opus 4.8 | [llm-stats](https://llm-stats.com/models/glm-5.3-flash) 9 | DeepSeek V4.1 Flash | $0.15 | deepseek-v4.1-flash | deepseek | 1.0M | 384K | low, high, max | 2025-05 | active | $0.15 | $0.60 | Excellent | 173,330 | 58 | 58 | GPT-6 Luna, Muse Spark 1.2 | [llm-stats](https://llm-stats.com/models/deepseek-v4.1-flash) 13 | GPT-5.6 Luna | $0.20 | gpt-5.6-luna | openai | 1.1M | 128K | none, low, medium, high, xhigh, max | 2026-02-16 | active | $0.20 | $1.20 | Moderate | 10,250 | 75 | 75 | — | [llm-stats](https://llm-stats.com/models/gpt-5.6-luna) 14 | MiniMax-M3 | $0.30 | minimax-m3 | minimax | 1.0M | 131K | — | 2025-01 | active | $0.30 | $1.20 | Moderate | 16,000 | 49 | 53 | Claude Haiku 5.5, Gemini 3.5 Flash | [llm-stats](https://llm-stats.com/models/minimax-m3) 14 | LongCat-2.0 | $0.30 | longcat-2.0 | meituan | 1.0M | 131K | — | — | active | $0.30 | $1.20 | Good | 57,200 | 36 | 39 | — | [llm-stats](https://llm-stats.com/models/longcat-2.0) 16 | Qwen3.7 Plus | $0.40 | qwen3.7-plus | qwen | 1M | 66K | — | 2025-04 | active | $0.40 | $1.60 | Good | 21,600 | 43 | 48 | — | [llm-stats](https://llm-stats.com/models/qwen3.7-plus) 17 | MiMo-V2.6-Pro | $0.43 | mimo-v2.6-pro | xiaomi | 1.1M | 131K | — | — | active | $0.43 | $0.87 | Moderate | 16,300 | 48 | 53 | Claude Haiku 5.5, Grok 4.7 | [llm-stats](https://llm-stats.com/models/mimo-v2.6-pro) 17 | MiMo V2.5 Pro | $0.43 | mimo-v2.5-pro | xiaomi | 1.1M | 128K | — | 2024-12 | active | $0.43 | $0.87 | Moderate | 16,300 | 66 | 66 | Claude Haiku 5.5, Grok 4.3 | [llm-stats](https://llm-stats.com/models/mimo-v2.5-pro) 19 | DeepSeek V4 Pro (New) | $0.66 | deepseek-v4-pro | deepseek | 1.0M | 384K | high, max | 2025-05 | active | $0.66 | $1.98 | High | 5,200 | 56 | 60 | Gemini 3.6 Flash, GPT-5.6 Luna | [llm-stats](https://llm-stats.com/models/deepseek-v4-pro-0813) 20 | Hy4 preview | $0.83 | hy4-preview | tencent | 1.0M | 64K | none, high | — | active | $0.83 | $2.50 | Moderate | 6,770 | 60 | 60 | Claude Haiku 5.5, MAI-Code-1.1-Flash | [llm-stats](https://llm-stats.com/models/hy4-preview) 21 | Kimi K2.7 Code | $0.95 | kimi-k2.7-code | moonshotai | 262K | 262K | — | 2025-01 | active | $0.95 | $4.00 | Moderate | 6,750 | 45 | 26 | Claude Haiku 5.5, Grok 4.3 | [llm-stats](https://llm-stats.com/models/kimi-k2.7-code) 22 | Step 5 Preview | $1.00 | step-5-preview | stepfun | 1.0M | 66K | low, medium, high | — | active | $1.00 | $2.70 | High | 2,130 | 68 | 68 | — | [llm-stats](https://llm-stats.com/models/step-5-preview) 23 | GLM-5.3 | $1.40 | glm-5.3 | z-ai | 1.0M | 131K | low, high, max | — | active | $1.40 | $4.40 | Extreme | 1,080 | 58 | 64 | GPT-5.5, Grok 4.6 | [llm-stats](https://llm-stats.com/models/glm-5.3) 23 | GLM-5.2 | $1.40 | glm-5.2 | z-ai | 1.0M | 131K | high, max | — | active | $1.40 | $4.40 | High | 4,300 | 72 | 72 | Gemini 3.6 Flash, Muse Spark 1.1 | [llm-stats](https://llm-stats.com/models/glm-5.2) 25 | Grok 4.7 | $2.00 | grok-4.7 | x-ai | 500K | 500K | low, medium, high, xhigh | 2026-05 | active | $2.00 | $6.00 | Extreme | 845 | 47 | 35 | — | [llm-stats](https://llm-stats.com/models/grok-4.7) 25 | Qwen3.8 Max | $2.00 | qwen3.8-max | qwen | 1M | 131K | low, medium, xhigh | — | active | $2.00 | $6.00 | Extreme | 810 | 53 | 60 | — | [llm-stats](https://llm-stats.com/models/qwen3.8-max) 27 | Kimi K3 | $3.00 | kimi-k3 | moonshotai | 1.0M | 131K | max | — | active | $3.00 | $15.00 | Extreme | 490 | 55 | 62 | Gemini 3.8 Flash, Claude Haiku 5.5 | [llm-stats](https://llm-stats.com/models/kimi-k3) 28 | Gemini 4 Argon Contributor | $4.00 | gemini-4-argon-contributor | google | 1.0M | 66K | low, medium, high | 2026-03 | active | $4.00 | $20.00 | High | 2,120 | 52 | 56 | — | [llm-stats](https://llm-stats.com/models/gemini-4-argon) - **Context** — maximum input context window in tokens. - **Output cap** — maximum output tokens (`—`: not published). - **Reasoning efforts** — reasoning-effort levels the model accepts (e.g. `low, high`); `—` if no effort control is published. - **Knowledge cutoff** — training-data cutoff date (`—`: not published). - **Status** — `active`, or `deprecated` when the upstream catalog marks the model deprecated. `—` means models.dev published no status for the model (or its catalog was unavailable), so the lifecycle is unknown rather than confirmed active. Advisory: a deprecated model may still be served. - **Burn** — how fast a model consumes the Go quota (`Excellent` is quota-friendly, `Extreme` burns fastest). - **Req/mo** — estimated requests per month at the $60 Go tier. - **Coding / Agentic** — task-fit score, 0–100 (algorithmic, not a benchmark). - **Replaces** — closed-source frontier models this Go model substitutes. - **Source** — LLM Stats page for the model. ## Query URLs The interactive table keeps its search and leaderboard in the URL, so every answer above has a link: - Leaderboard: `https://zp.svelte-apps.me/?need=`, where `` is one of `coding`, `design`, `smartest`, `reasoning`, `vision`, `open-source`, `long-context`, `cheapest`. - Search: `https://zp.svelte-apps.me/?q=` across model name, provider, tags, and capabilities. - Both at once: `https://zp.svelte-apps.me/?need=cheapest&q=vision`. - The same view as data: `https://zp.svelte-apps.me/llms.txt?need=` or `https://zp.svelte-apps.me/llms.txt?q=` returns this file with the model table ranked or filtered, using the site's own algorithm. - Side by side: `https://zp.svelte-apps.me/compare?models=,` (up to four model IDs). - Data, not prose: `https://zp.svelte-apps.me/debug/export?section=&format=json` (or `format=md`) returns one slice of the catalog with the right content type. Sections: `overview`, `pricing`, `quota`, `burn`, `benchmarks`, `matching`, `lifecycle`, `migration`, `compare`, `tags`, `scenarios`, `raw`. Those URLs render in the browser, so a crawler that cannot run JavaScript sees only the page shell; the markdown in this file is the machine-readable form of the same data. ## Key pages - [Model comparison](https://zp.svelte-apps.me/): sortable table, scenario filters, and quota calculator - [Side-by-side comparison](https://zp.svelte-apps.me/compare): compare up to four models on pricing, benchmarks, and fit - [Methodology & about](https://zp.svelte-apps.me/about): how scores, burn bands, and migration hints are computed ## Data sources - [modelgrep](https://modelgrep.com): OpenRouter pricing plus Artificial Analysis and Design Arena benchmarks - [LLM Stats](https://llm-stats.com): model benchmarks and release metadata - [OpenCode Go docs](https://opencode.ai/v2/docs/console/go): official model list, pricing, and usage limits - [models.dev](https://models.dev): first-party OpenCode catalog metadata (display names, output caps, reasoning efforts, knowledge cutoff) and the Go model lifecycle history. Advisory only: it never changes which models are listed. ## Notes - This file is generated on request at `https://zp.svelte-apps.me/llms.txt` from ZenPick's server-side model cache. - Catalog data refreshes automatically (stale-while-revalidate, ~6h TTL). - `?need=` and `?q=` change the model table in this file, so caches should key on the full URL. - Machine operators can force a fresh upstream fetch via `POST /api/revalidate` (see repository README).