update-opencode-go-models

name	update-opencode-go-models
description	Update OpenCode Go model rate limits and availability in chezmoi/dot_pi/agent/models.json

Update OpenCode Go Models

Document status: Active

Overview

Keep chezmoi/dot_pi/agent/models.json in sync with OpenCode Go's live model catalog, rate limits, and Artificial Analysis coding index annotations.

Scope: Only the providers.opencode-go.modelOverrides object. This skill does not apply to the openrouter provider or the xai provider.

Sources of Truth

Source	URL	Notes
Rate limit docs	gh api repos/anomalyco/opencode/contents/packages/web/src/content/docs/go.mdx --jq '.content' | base64 -d	⚠️ raw.githubusercontent.com truncates output at ~180 lines. Use gh api instead.
Model IDs	curl -s https://models.dev/api.json	jq parsing may fail; save to file first
AA coding index	curl -s https://artificialanalysis.ai/api/v2/data/llms/models -H "x-api-key: $ARTIFICIAL_ANALYSIS_API_KEY" -o /var/tmp/aa.json	Requires $ARTIFICIAL_ANALYSIS_API_KEY to be set. Save to /var/tmp/aa.json.

Target file to modify

chezmoi/dot_pi/agent/models.json — the providers.opencode-go.modelOverrides object.

Rate Limit Formula

All name fields must follow this exact pattern:

<Display Name> (reqs: <N>/hr <M>/min)

Where:

• <N> = requests per 5 hours ÷ 5 (round to nearest integer)
• <M> = requests per 5 hours ÷ 5 ÷ 60 (round to 1 decimal place)

⚠️ Common mistake: Do NOT calculate per-minute as <N> ÷ 60. Calculate both per-hour and per-minute from the original requests per 5 hours value.

Analytics Annotation Formula

For models with a non-null AA coding index, append an analytics suffix to the name field:

<Display Name> (reqs: <N>/hr <M>/min) ⟐  <coding> · ●  <product>k

Where:

• <coding> = Artificial Analysis Coding Index (1 decimal place, e.g. 68.8)
• <product> = requests_per_month × coding_index ÷ 1000, rounded to nearest integer, displayed as k-suffixed whole number (e.g. 296k, 8888k)
• If a model has no AA coding index (null), the analytics suffix is omitted entirely — the name stays as the rate-limit-only format

Examples

Model	Name field
DeepSeek V4 Flash	DeepSeek V4 Flash (reqs: 6330/hr 105.5/min) ⟐ 56.2 · ● 8888k
GLM-5.2	GLM-5.2 (reqs: 176/hr 2.9/min) ⟐ 68.8 · ● 296k
MiMo-V2.5	MiMo-V2.5 (reqs: 6020/hr 100.3/min)

AA Slug Mapping to OpenCode Go Model IDs

AA uses hyphenated slugs. Map from AA slug to OpenCode Go model ID:

AA Slug	OpenCode Go Model ID
deepseek-v4-pro	deepseek-v4-pro
deepseek-v4-flash	deepseek-v4-flash
glm-5-2	glm-5.2
glm-5-1	glm-5.1
kimi-k2-7-code	kimi-k2.7-code
kimi-k2-6	kimi-k2.6
mimo-v2-5-pro	mimo-v2.5-pro
mimo-v2-5-0424	mimo-v2.5
minimax-m3	minimax-m3
minimax-m2-7	minimax-m2.7
qwen3-7-max	qwen3.7-max
qwen3-7-plus	qwen3.7-plus
qwen3-6-plus	qwen3.6-plus

Rate Limit Reference Table

Extract from go.mdx section "Usage limits". Use these values for calculations:

Model	Requests per 5hr	Per Hour (÷5)	Per Min (÷5÷60)
GLM-5.2	880	176	2.9
GLM-5.1	880	176	2.9
Kimi K2.6	1150	230	3.8
Kimi K2.7 Code	1350	270	4.5
MiMo-V2.5	30100	6020	100.3
MiMo-V2.5-Pro	3250	650	10.8
MiniMax M3	3200	640	10.7
MiniMax M2.7	3400	680	11.3
Qwen3.7 Max	950	190	3.2
Qwen3.7 Plus	4300	860	14.3
Qwen3.6 Plus	3300	660	11.0
DeepSeek V4 Pro	3450	690	11.5
DeepSeek V4 Flash	31650	6330	105.5

Model ID Mapping

Display names in the docs vs kebab-case model IDs:

Display Name	Model ID
GLM-5.2	glm-5.2
GLM-5.1	glm-5.1
Kimi K2.6	kimi-k2.6
Kimi K2.7 Code	kimi-k2.7-code
MiMo-V2.5	mimo-v2.5
MiMo-V2.5-Pro	mimo-v2.5-pro
MiniMax M3	minimax-m3
MiniMax M2.7	minimax-m2.7
Qwen3.7 Max	qwen3.7-max
Qwen3.7 Plus	qwen3.7-plus
Qwen3.6 Plus	qwen3.6-plus
DeepSeek V4 Pro	deepseek-v4-pro
DeepSeek V4 Flash	deepseek-v4-flash

Process

┌─────────────────────────────┐
│ 1. Fetch full go.mdx docs   │
│    (use gh api method)      │
└────────────┬────────────────┘
             │
             ▼
┌─────────────────────────────┐
│ 2. Fetch AA model data      │
│    (curl → /var/tmp/aa.json)│
└────────────┬────────────────┘
             │
             ▼
┌─────────────────────────────┐
│ 3. Extract rate table       │
│    from "Usage limits"      │
└────────────┬────────────────┘
             │
             ▼
┌─────────────────────────────┐
│ 4. Extract AA coding indices│
│    per model slug           │
└────────────┬────────────────┘
             │
             ▼
┌─────────────────────────────┐
│ 5. Calculate per-hr/min,    │
│    coding product, build    │
│    name strings             │
└────────────┬────────────────┘
             │
             ▼
┌─────────────────────────────┐
│ 6. Compare with current     │
│    JSON and update          │
└────────────┬────────────────┘
             │
             ▼
┌─────────────────────────────┐
│ 7. Validate output          │
│    JSON is valid            │
└────────────┬────────────────┘
             │
             ▼
┌─────────────────────────────┐
│ 8. Commit                   │
└─────────────────────────────┘

Python Calculation Script

Save to /tmp/calc_ratelimits.py and run python3 /tmp/calc_ratelimits.py:

#!/usr/bin/env python3
"""Recalculate OpenCode Go model rate limits + analytics from docs and AA data."""

import json

# ── Rate limit + requests-per-month data (from go.mdx) ──────────
MODELS = {
    "GLM-5.2":        {"reqs_5hr": 880,   "reqs_mo": 4300},
    "GLM-5.1":        {"reqs_5hr": 880,   "reqs_mo": 4300},
    "Kimi K2.6":      {"reqs_5hr": 1150,  "reqs_mo": 5750},
    "Kimi K2.7 Code":  {"reqs_5hr": 1350,  "reqs_mo": 9250},
    "MiMo-V2.5":      {"reqs_5hr": 30100, "reqs_mo": 150400},
    "MiMo-V2.5-Pro":  {"reqs_5hr": 3250,  "reqs_mo": 16300},
    "MiniMax M3":     {"reqs_5hr": 3200,  "reqs_mo": 16000},
    "MiniMax M2.7":   {"reqs_5hr": 3400,  "reqs_mo": 17000},
    "Qwen3.7 Max":    {"reqs_5hr": 950,   "reqs_mo": 4770},
    "Qwen3.7 Plus":   {"reqs_5hr": 4300,  "reqs_mo": 21600},
    "Qwen3.6 Plus":   {"reqs_5hr": 3300,  "reqs_mo": 16300},
    "DeepSeek V4 Pro":   {"reqs_5hr": 3450,  "reqs_mo": 17150},
    "DeepSeek V4 Flash": {"reqs_5hr": 31650, "reqs_mo": 158150},
}

DISPLAY_TO_ID = {
    "GLM-5.2": "glm-5.2",
    "GLM-5.1": "glm-5.1",
    "Kimi K2.6": "kimi-k2.6",
    "Kimi K2.7 Code": "kimi-k2.7-code",
    "MiMo-V2.5": "mimo-v2.5",
    "MiMo-V2.5-Pro": "mimo-v2.5-pro",
    "MiniMax M3": "minimax-m3",
    "MiniMax M2.7": "minimax-m2.7",
    "Qwen3.7 Max": "qwen3.7-max",
    "Qwen3.7 Plus": "qwen3.7-plus",
    "Qwen3.6 Plus": "qwen3.6-plus",
    "DeepSeek V4 Pro": "deepseek-v4-pro",
    "DeepSeek V4 Flash": "deepseek-v4-flash",
}

# ── AA slug → model ID mapping ─────────────────────────────────
AA_SLUG_TO_MODEL_ID = {
    "deepseek-v4-pro":   "deepseek-v4-pro",
    "deepseek-v4-flash": "deepseek-v4-flash",
    "glm-5-2":          "glm-5.2",
    "glm-5-1":          "glm-5.1",
    "kimi-k2-7-code":   "kimi-k2.7-code",
    "kimi-k2-6":        "kimi-k2.6",
    "mimo-v2-5-pro":    "mimo-v2.5-pro",
    "mimo-v2-5-0424":   "mimo-v2.5",
    "minimax-m3":       "minimax-m3",
    "minimax-m2-7":     "minimax-m2.7",
    "qwen3-7-max":      "qwen3.7-max",
    "qwen3-7-plus":     "qwen3.7-plus",
    "qwen3-6-plus":     "qwen3.6-plus",
}

def load_aa_coding(path="/var/tmp/aa.json"):
    """Load AA coding indices, return dict of model_id → coding or None."""
    with open(path) as f:
        data = json.load(f)
    coding = {}
    for entry in data["data"]:
        aa_slug = entry["slug"]
        if aa_slug in AA_SLUG_TO_MODEL_ID:
            model_id = AA_SLUG_TO_MODEL_ID[aa_slug]
            raw = entry["evaluations"]["artificial_analysis_coding_index"]
            coding[model_id] = round(raw, 1) if raw is not None else None
    return coding

def per_hour(reqs_5hr):
    return round(reqs_5hr / 5)

def per_minute(reqs_5hr):
    return round(reqs_5hr / 5 / 60, 1)

def product_k(reqs_mo, coding):
    """req×coding ÷ 1000, rounded to nearest integer."""
    if coding is None:
        return None
    return round(reqs_mo * coding / 1000)

def full_name(display_name, reqs_5hr, reqs_mo, coding):
    """Build the full name string with rate limits and optional analytics suffix."""
    pH = per_hour(reqs_5hr)
    pM = per_minute(reqs_5hr)
    base = f"{display_name} (reqs: {pH}/hr {pM}/min)"
    if coding is not None:
        pk = product_k(reqs_mo, coding)
        return f"{base} \u27d0  {coding} \u00b7 \u25cf  {pk}k"
    return base

def main():
    aa_coding = load_aa_coding()

    print("Model ID | Display Name | Reqs/5hr | Per Hour | Per Min | Coding | Product(k)")
    print("---------|--------------|----------|----------|---------|--------|------------")
    for display_name, data in sorted(MODELS.items()):
        model_id = DISPLAY_TO_ID[display_name]
        coding = aa_coding.get(model_id)
        pH = per_hour(data["reqs_5hr"])
        pM = per_minute(data["reqs_5hr"])
        pk = product_k(data["reqs_mo"], coding)
        coding_str = str(coding) if coding is not None else "N/A"
        pk_str = str(pk) if pk is not None else "N/A"
        print(f'"{model_id}": {{"name": "{full_name(display_name, data["reqs_5hr"], data["reqs_mo"], coding)}"}}')
        print(f"  Rate: {data['reqs_5hr']}/5hr → {pH}/hr {pM}/min | AA: {coding_str} | xP: {pk_str}k")

    print("\n--- JSON snippet ---")
    entries = []
    for display_name, data in sorted(MODELS.items()):
        model_id = DISPLAY_TO_ID[display_name]
        coding = aa_coding.get(model_id)
        entries.append(f'        "{model_id}": {{"name": "{full_name(display_name, data["reqs_5hr"], data["reqs_mo"], coding)}"}}')

    print('"opencode-go": {\n      "modelOverrides": {\n' + ",\n".join(entries) + "\n      }\n    },")

if __name__ == "__main__":
    main()

Common Mistakes to Avoid

1. Fetching truncated docs: curl https://raw.githubusercontent.com/... returns ~32 lines. Use gh api ... | base64 -d to get the full 178 lines.

2. Wrong per-minute calculation: Per-minute is reqs_5hr / 5 / 60, NOT per_hour / 60. Calculate both values from the original requests per 5 hours value.

3. Swapped rate limits: Verify that the cheaper model (Qwen3.5 Plus) has MORE requests/hr than the expensive one (Qwen3.6 Plus). Qwen3.5 Plus should be ~2040/hr, Qwen3.6 Plus should be ~660/hr.

4. Inconsistent naming: Use exact display names from docs (e.g., GLM-5.1 not GLM 5.1).

5. Missing AA data fetch: If AA API returns an error or empty data, skip the analytics suffix for all models rather than failing. Log the error, proceed with rate-limit-only names.

6. Stale AA data in cache: Always re-fetch /var/tmp/aa.json each run. Do not reuse a cached copy.

7. Wrong AA slug: AA slugs use hyphens (e.g. glm-5-2, mimo-v2-5-0424), while OpenCode Go model IDs use dots (e.g. glm-5.2). The AA_SLUG_TO_MODEL_ID mapping handles this — do not attempt to derive one from the other algorithmically.