Repository navigation
No usage/cost shown for DeepInfra provider (openai-compatible) #45483
Description
Description
When using DeepInfra (https://deepinfra.com/) as a provider, opencode shows no token usage and no cost in the TUI / stats / session sidebar. The same setup works correctly with Anthropic (Claude), DeepSeek and OpenRouter — usage and cost display fine there. DeepInfra is the only one that shows nothing.
Setup
DeepInfra is configured as an OpenAI-compatible provider (@ai-sdk/openai-compatible), baseURL https://api.deepinfra.com/v1, e.g.:
"provider": {
"deepinfra": {
"npm": "@ai-sdk/openai-compatible",
"options": { "baseURL": "https://api.deepinfra.com/v1" },
"models": { "deepseek-ai/DeepSeek-V4-Flash-0731": { "name": "DS Flash DI", "tool_call": true, "reasoning": true } }
}
}Expected
Usage (prompt/completion tokens, cached tokens) and cost are reported for the DeepInfra provider, like they are for OpenRouter/DeepSeek.
Actual
No usage / cost is shown for DeepInfra at all.
Environment
- opencode (OpenChamber UI)
- macOS
- Provider: DeepInfra via OpenAI-compatible adapter
Related
- No token usage tracking with @ai-sdk/openai-compatible provider #41328 No token usage tracking with @ai-sdk/openai-compatible provider
- OpenAI-compatible cost chunk omits required stream fields #39061 OpenAI-compatible cost chunk omits required stream fields
- [FEATURE] Honor provider-reported cost (usage.cost) from LLM gateways (OpenRouter, LiteLLM, Manifest) #43818 Honor provider-reported cost from LLM gateways
Activity
Tobias-Conrad commentedon Sep 16, 2026
Cross-reference: the DeepInfra account balance now shows in OpenChamber's Usage panel (openchamber/openchamber#3600) via
GET https://api.deepinfra.com/v1/me?checklist=true→checklist.stripe_balance. That covers the balance surface; the per-request token/cost reporting for@ai-sdk/openai-compatibleproviders in the TUI is separate and still open here.opencodejr commentedon Oct 28, 2026
Thanks for the report! A few details would help us look into this:
opencode --version.opencodeTUI (session sidebar) or withopencode stats? If it only happens in OpenChamber, the problem is probably there.deepinframodel has nocostentry, so its cost will show as $0 even when tokens are tracked. You can add the rates to the model config, for example:You can reply here with the details, or edit the issue. If you edit it, please also leave a comment saying you've updated it, so we know you're done.
Related issues I found, ordered from most to least likely to be relevant:
@ai-sdk/openai-compatible(closed because the issue body was empty)usage.cost) instead of computing it locallyopencodejr commentedon Oct 31, 2026
Closing this for now because we haven't heard back. To look into the missing DeepInfra usage and cost, we'd need your OpenCode version (
opencode --version). We'd also need to know whether it happens in the plainopencodeTUI oropencode stats, or only in OpenChamber, and whether the token counts or only the cost are missing.A quick tip in the meantime: OpenCode works out cost from the model's per-token rates. A custom model with no
costentry in its config will always show $0. Adding"cost": { "input": ..., "output": ..., "cache_read": ... }(USD per 1M tokens) to the model may be enough.If it still happens, please comment here with those details or open a new issue, and we'll take another look. Thanks!