How do I use DeepSeek V4 Pro in OpenCode or Claude Code without per-token billing?
Use the model ID deepseek/deepseek-v4-pro-0813 on a Standard Compute plan from $19: that is the August release DeepSeek's own API calls deepseek-v4-pro, and every plan carries it on a monthly compute budget that stops new requests with HTTP 402 at 100%. The April original stays available as deepseek/deepseek-v4-pro.
| Model | The larger model in DeepSeek's open-weight V4 line, with a 1M-token context window. The current version, 0813, dates from August 13, 2026. |
|---|---|
| Model IDs | deepseek/deepseek-v4-pro-0813April original: deepseek/deepseek-v4-pro |
| Base URL | https://api.stdcmpt.com/v1Claude Code uses https://api.stdcmpt.com |
| Input | Text only |
| Reasoning | 0813: minimal, low, medium or high. April original: on or off |
| Model ID, US pin | deepseek/deepseek-v4-pro-0813:usApril original: deepseek/deepseek-v4-pro:us |
Copy a working config
OpenCode, Claude Code and Kilo Code all take the dated ID. Request Logs show which model answered each call, so you can confirm you got 0813.
{
"$schema": "https://opencode.ai/config.json",
"model": "standardcompute/deepseek/deepseek-v4-pro-0813",
"provider": {
"standardcompute": {
"npm": "@ai-sdk/openai-compatible",
"name": "Standard Compute",
"options": {
"baseURL": "https://api.stdcmpt.com/v1",
"apiKey": "{env:STANDARDCOMPUTE_API_KEY}"
},
"models": {
"deepseek/deepseek-v4-pro-0813": { "name": "DeepSeek V4 Pro 0813" },
"standardcompute": { "name": "Smart routing" }
}
}
}
}
Replace YOUR_API_KEY with a key from the dashboard. This config reads your key from STANDARDCOMPUTE_API_KEY; export it with a key from the dashboard. Full setup: OpenCode, Claude Code, Kilo Code.
What it costs
| Per 1M tokens | Peak | Off-peak |
|---|---|---|
| Input, cache miss | $1.32 | $0.66 |
| Input, cache hit | $0.044 | $0.022 |
| Output | $3.96 | $1.98 |
DeepSeek API rates for deepseek-v4-pro (V4 Pro 0813). Peak is 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday; every other hour is off-peak. Source: DeepSeek API pricing, checked October 13, 2026.
Average coding-agent turn on our gateway, Sep 13 to Oct 12, 2026. DeepSeek peak list prices; off-peak, the same turn costs half.
Cache hits are the cheap part: on the turn above, 87% of the prompt costs about a sixth of the bill, so output and fresh input decide what a V4 Pro session costs.
What works differently here
- The short form
deepseek-v4-proresolves to the April original here, not to 0813. DeepSeek's API uses that name for 0813, so copy the dated IDdeepseek/deepseek-v4-pro-0813to get the model DeepSeek serves today. - DeepSeek V4.1 Flash, released a month later, scores higher on Artificial Analysis's Intelligence Index (39 against 36 for V4 Pro 0813) and lists at less than a quarter of V4 Pro's peak input price. Try it before pinning Pro.
- Paying DeepSeek directly costs half off-peak. A V4 Pro-only workload that runs mostly at night or on weekends can cost less there.
Other ways to run DeepSeek V4 Pro
| DeepSeek API$1.32 / $3.96 per 1M | First-party as deepseek-v4-pro, half price off-peak. |
|---|---|
| OpenRouter hostsPay per token | Several hosts serve both versions at their own rates and speeds. |
| Self-hostGPU + power | Open weights; at V4 Pro's size it needs multi-GPU server hardware. |
| Standard ComputeMonthly plan | Both V4 Pro versions and both Flash generations, with US pins for Pro. |
FAQ
Which ID is DeepSeek V4 Pro 0813?
deepseek/deepseek-v4-pro-0813. The undated deepseek/deepseek-v4-pro is the April original. Both are on every plan; GET https://api.stdcmpt.com/v1/models lists them.
Does DeepSeek V4 Pro read images?
No, both versions are text-only. Select a vision model next to it, such as DeepSeek V4.1 Flash, and image requests go to that model.
DeepSeek V4 Pro or GLM 5.3?
Both are text-only open-weight models with 1M context. DeepSeek V4 Pro 0813 lists at $1.32 / $3.96 per million tokens at peak and half that off-peak; Z.ai lists GLM 5.3 at $1.40 / $4.40, so off-peak DeepSeek is the cheaper of the two.
Can I control how long V4 Pro thinks?
On 0813, yes: minimal, low, medium or high, per request or as a Models page default. The April original only switches reasoning on or off.
Sources: DeepSeek API pricing; Artificial Analysis: DeepSeek V4 Pro; Artificial Analysis: DeepSeek V4.1 Flash
Plans from $19/mo
One key for DeepSeek V4 Pro and the rest of the catalog, with a monthly compute budget you can watch in the dashboard and no overage charges.