OpenCode Go limits, measured
OpenCode Go’s limits are dollar amounts per model: the 5-hour limit is 20% of the monthly limit and the weekly limit 50%. OpenCode also publishes request estimates, which assume a request that is 99% cache hits. Real coding traffic is not that cached: DeepSeek traffic through Wokey is 81.3% cache hits and 16.9% uncached input. At that mix a Go 5-hour window of DeepSeek V4.1 Flash is about 520M tokens, roughly 1/6 of the tokens the official estimate implies.
- Official dollar limits
- Real token mix
- Split by model
- Go · 5 hours · DeepSeek V4.1 Flash
- 520M / 260M · off-peak / peak, measured mix
- Requests in that window
- about 13,000 · OpenCode estimates 43,330; real requests average 41K tokens
- Weekly limit
- 2.5 windows · 5 a month; counted per model
- Measured cache share
- 81.3% · 34 accounts, 2.28B tokens
Dollar limits by model (official, Go plan)
Each model has its own monthly limit; the 5-hour limit is 20% of it and the weekly limit 50%. Request counts are OpenCode’s estimates for “typical usage”; the last two columns are the tokens those estimates correspond to and the cache share they assume.
| Model | Per month | Per 5 hours | Per week | Estimated requests / 5 hours | Tokens / 5 hours | Assumed cache share |
|---|---|---|---|---|---|---|
| DeepSeek V4.1 Flash | $100 | $20 | $50 | 43,330 | 3.1B | 99% |
| GLM-5.3-Flash | $60 | $12 | $30 | 6,320 | 350M | 97.9% |
| MiniMax M3 | $60 | $12 | $30 | 3,200 | 180M | 98.8% |
| Kimi K3.1 Pro | $15 | $3 | $7.5 | 147 | 11M | 98.3% |
OpenCode Go has about 30 models; these are the ones with the most traffic through Wokey. Token prices are the same on all three plans. Free models do not count.
DeepSeek V4.1 Flash: what a window really holds
Go allows $20 per 5 hours. DeepSeek costs $0.15 input, $0.6 output and $0.003 cache read per million tokens, doubled at peak (01:00–04:00 and 06:00–10:00 UTC, Monday to Friday). Uncached input costs 50× a cache read, so a slightly lower cache share means far fewer tokens.
| Mix | Cost per 1M tokens | Tokens per 5 hours | Requests per 5 hours | Tokens per month |
|---|---|---|---|---|
| The request OpenCode’s estimate assumes (99% cache) | $0.0064 | 3.1B | 43,330 | 16B |
| Measured mix (81.3% cache) · off-peak | $0.039 | 520M | 13,000 | 2.6B |
| Measured mix (81.3% cache) · peak | $0.077 | 260M | 6,400 | 1.3B |
The measured mix is DeepSeek traffic in 74 stretches of a 5-hour window on 34 Go accounts: 16.9% uncached input, 81.3% cache hits, 1.8% output, 40,600 tokens per request on average. Go Plus has twice Go’s DeepSeek limit. Past the limit, DeepSeek V4.1 Flash through Wokey costs $0.075 input / $0.3 output per million tokens.
Monthly limits on the three plans (official)
Prices are per month. Each column is that model’s monthly dollar limit; the 5-hour and weekly limits are 20% and 50% of it.
| Plan | Per month | DeepSeek V4.1 Flash | GLM-5.3-Flash | MiniMax M3 | Kimi K3.1 Pro |
|---|---|---|---|---|---|
| Go | $10 | $100 | $60 | $60 | $15 |
| Go Plus | $40 | $200 | $180 | $180 | $60 |
| Go Teams | $20 per member | $105 | $65 | $65 | $20 |
Go Plus costs 4× Go but has only 2× the DeepSeek V4.1 Flash limit, 3× for GLM-5.3-Flash and MiniMax M3, and 4× for Kimi K3.1 Pro. If you mostly use DeepSeek, Plus is poor value.
Limit rules
- Each model has three limits: 5 hours (20% of the monthly limit), weekly (50%) and monthly (100%).
- Limits are in dollars: each request is converted at the model’s token prices and counted against that model.
- DeepSeek costs double at peak hours, so the same tokens use twice the limit.
- At a limit, requests are blocked by default and free models still work; with Extra Usage on, usage falls back to your pay-as-you-go balance.
- Some models are restricted by region or privacy settings; DeepSeek needs a Global region.
How to stretch the limit
- Keep the cache above all: uncached DeepSeek input costs 50× a cache read. Avoid switching models or changing the system prompt and tool list, and do not restart long sessions needlessly.
- On DeepSeek, avoid 01:00–04:00 and 06:00–10:00 UTC on weekdays, when prices double.
- Limits are per model: when one runs out you can continue on another.
- Kimi K3.1 Pro gets only $3 per 5 hours on Go, about 147 requests by OpenCode’s estimate; keep it for occasional use.
Method and limits
- Dollar limits, token prices, request estimates and the request size they assume are from OpenCode’s documentation, unchanged.
- Measured mix: DeepSeek traffic through Wokey in 74 stretches of a 5-hour window on 34 Go accounts from 2026-11-07 to 11-14, 2.28B tokens. Wokey’s traffic comes from many users and clients, so its cache share may be lower than one user’s long session in OpenCode.
- Reconciliation: in the 26 stretches that were fully off-peak, usage rose 1.25–2.3× what list prices predict, median 1.6×. Readings are whole percentages and account owners also use their accounts, so the gap does not show overcharging; it only shows no case of charging less than the formula.
- Only DeepSeek has enough real traffic. Other models show official figures only.
- OpenCode can change limits, prices and the model list at any time; the date on this page is the last check.
FAQ
What are the OpenCode Go limits?
Dollar limits per model. On Go ($10 a month) DeepSeek V4.1 Flash gets $100 a month, $50 a week and $20 per 5 hours; GLM-5.3-Flash and MiniMax M3 get $60 a month; Kimi K3.1 Pro $15. The 5-hour limit is 20% of the monthly limit and the weekly limit 50%.
How many tokens fit in 5 hours on OpenCode Go?
It depends on the model and the cache share. DeepSeek V4.1 Flash gets $20 per 5 hours on Go: at the mix Wokey measured (81.3% cache) that is about 520M tokens and 13,000 requests off-peak, and 260M at peak. OpenCode’s estimate of 43,330 requests corresponds to about 3.1B tokens, if 99% hit the cache.
Why do I get fewer requests than OpenCode’s estimate?
The estimate assumes a request with 71,300 cached tokens and only 410 uncached input and 310 output tokens. Real traffic is 16.9% uncached input, which costs 50× a cache read, so a million tokens cost about 6× the estimate. Peak hours double it again.
How do Go and Go Plus differ?
Go Plus costs $40 a month, 4× Go. Limits do not all scale by 4: DeepSeek V4.1 Flash is 2× ($200 a month), GLM-5.3-Flash and MiniMax M3 3× ($180) and Kimi K3.1 Pro 4× ($60).
When do OpenCode Go limits reset?
There are 5-hour, weekly and monthly limits, counted per model. The weekly limit is 2.5× the 5-hour limit and the monthly limit 5×, so at most 2.5 full 5-hour windows a week and 5 a month.
What happens when I hit the limit?
Requests are blocked by default. You can use the free models, switch to a model that still has room (limits are per model), turn on Extra Usage to draw on your pay-as-you-go balance, or use a per-token API.
Is OpenCode Go worth it?
Go costs $10 a month and DeepSeek V4.1 Flash alone has a $100 monthly limit at list prices, with other models on top, far above the fee if you use it up. The constraint is the window: $20 per 5 hours, about 13,000 requests at the measured mix, so heavy use hits the 5-hour and weekly limits first.