Grok 4.5
FlagshipxAI flagship and one of the strongest open coding models; ~120 requests per 5h.
Back to comparisonOpenCode Go
OpenCode · Anomaly
OpenCode Go
Plan Guide
Low-cost subscription for open coding models: $5 for the first month, then $10/month with about 6x the subscription value in usage. A curated lineup of 18 tested open-source models works with OpenCode or any compatible agent.
18Models
$5First month
6xUsage value
31,650Top 5h requests
Models
18 open models, all tested and benchmarked for agentic coding. The lineup grows over time.
GLM-5.2
LMArena #1Zhipu flagship, #1 open-source on LMArena Code; ~880 requests per 5h.
Kimi K3
Moonshot flagship (2.8T params, 1M context); ~110 requests per 5h.
Qwen3.8 Max
Alibaba’s newest flagship; ~160 requests per 5h.
DeepSeek V4
2x flash promoV4 Pro ~3,450 requests/5h; V4 Flash ~31,650 with limited-time 2x limits.
MiniMax M3
Multimodal flagship; ~3,200 requests per 5h.
MiMo-V2.5
Best valueXiaomi’s ultra-cheap open model; ~30,100 requests per 5h.
Hy3
High-performance open model; ~4,300 requests per 5h.
GPT 5.6 Luna
OpenAI open model; ~2,050 requests per 5h.
Plans
A single $10/mo tier with 6x usage value. $5 first month. Top up credit if needed, cancel anytime.
Go
$5 first month$10/mo
$5 first month · Invite friends - both get $5
- $12 / 5h · $30 / week · $60 / month usage
- 18 curated open models (Grok 4.5, GLM-5.2, Kimi K3...)
- OpenAI / Anthropic compatible APIs, any agent
- Fall back to Zen balance when limits are reached
- No training on your data, 0-day retention for most models
- Cancel any time, no lock-in
Best for
Subscribe to Go →Developers who want the strongest open models at the lowest cost, and heavy agent users
Model usage
Estimated requests at typical usage patterns (per 5h / week / month).
| Model | 5 hours | Weekly | Monthly |
|---|---|---|---|
| DeepSeek V4 Flash | 31,650 | 79,050 | 158,150 |
| MiMo-V2.5 | 30,100 | 75,200 | 150,400 |
| Qwen3.7 Plus | 4,300 | 10,800 | 21,600 |
| Hy3 | 4,300 | 10,750 | 21,500 |
| DeepSeek V4 Pro | 3,450 | 8,550 | 17,150 |
| MiniMax M3 | 3,200 | 8,000 | 16,000 |
| GLM-5.2 | 880 | 2,150 | 4,300 |
| Qwen3.8 Max | 160 | 400 | 810 |
| Grok 4.5 | 120 | 300 | 600 |
| Kimi K3 | 110 | 250 | 490 |
支持的编程工具
OpenCodeClaude CodeCursorClineRoo CodeKilo CodeCherry Studio
Usage limits
Limits are dollar-based: $12 per 5h, $30 per week, $60 per month. Expensive models consume budget faster. Free models remain available after you hit limits.
Rolling 5-hour
$12 of usage per rolling 5-hour window.
Weekly
$30 per week, reset weekly, no rollover.
Monthly
$60 per month - about 6x your $10 subscription.
Balance fallback
Enable "Use balance" in the console to continue from your Zen balance after limits.
How to connect
- Sign in at opencode.ai, subscribe to Go and copy your API key
- In the OpenCode TUI run /connect, pick OpenCode Go, paste your key
- Model IDs use the format opencode-go/<model-id> (e.g. opencode-go/kimi-k3)
- OpenAI / Anthropic / Responses compatible endpoints work with Claude Code, Cursor and more
Privacy
- No model is used for training
- Most models have 0-day data retention (GLM, Kimi, Qwen, DeepSeek, MiniMax, MiMo)
- Grok 4.5 and GPT 5.6 Luna retain 30 days (abuse monitoring)
- DeepSeek V4 Flash zero-data-retention agreement renews monthly
Frequently Asked Questions
What is OpenCode Go?
OpenCode Go is Anomaly’s low-cost subscription for open coding models: $5 for the first month, then $10/month, giving about 6x the subscription value in usage across 18 curated open models, usable in OpenCode or any OpenAI/Anthropic compatible agent.
How is Go different from Zen?
Go is the low-cost open-model subscription ($10/mo, 6x value). Zen is OpenCode’s flagship subscription with premium closed-model credits. They are separate subscriptions - Go is the open-source model month pass.
How do usage limits work?
Limits are dollar-based: $12 per 5 hours, $30 per week, $60 per month. Request counts depend on model price - DeepSeek V4 Flash allows ~31,650 requests/5h while Kimi K3 allows ~110. After limits you can fall back to Zen balance or free models.
Which models are included?
18 models: Grok 4.5, GLM-5.2/5.1, GPT 5.6 Luna, Kimi K3/K2.7 Code/K2.6, MiMo-V2.5/-Pro, MiniMax M3/M2.7, Qwen3.8 Max/Qwen3.7 Max/Plus/Qwen3.6 Plus, DeepSeek V4 Pro/V4 Flash, Hy3. The list keeps growing.
Do I get a discount for referrals?
Yes. Friends who join and subscribe to Go through your invite link (https://opencode.ai/go?ref=JBT5KJRCD4) earn $5 usage credit for both of you.
Is my data used for training?
No model uses your data for training. Most models keep 0-day retention; Grok 4.5 and GPT 5.6 Luna keep 30 days of logs for abuse monitoring.