Price and latency
Every provider and model AI called in the last 7 UTC days (2026-10-15 to 2026-10-22). Latency is the upper bound of a fixed latency bucket, the slowest day in the window; prices are per million tokens from the current rate cards.
| Provider | Model | Capability | p50 | p95 | Success rate | Price | Free allowance |
|---|---|---|---|---|---|---|---|
local | qwen2.5-1.5b-instruct | chat, generate, summarize, classify, extract | 25,600 ms | 25,600 ms | 100 % (10 / 10) | not billed | — |
shared | gpt-5-mini | chat | 25,600 ms | 25,600 ms | 0 % (0 / 3) | in $0.2500 · cached $0.0250 · out $2.0000 per 1M tokens | — |
shared | gpt-5-mini-2025-08-07 | chat, generate, summarize, classify, extract, enrich, json, vision, embed, transcribe, tools | 6,400 ms | 25,600 ms | 100 % (951 / 951) | in $0.2500 · cached $0.0250 · out $2.0000 per 1M tokens | — |
shared | gpt-5-nano | chat | 25,600 ms | 25,600 ms | 0 % (0 / 3) | in $0.0500 · cached $0.005000 · out $0.4000 per 1M tokens | — |
shared | gpt-5-nano-2025-08-07 | chat, generate, summarize, classify, extract, enrich, json, vision, embed, transcribe, tools | 3,200 ms | 25,600 ms | 100 % (9,233 / 9,233) | in $0.0500 · cached $0.005000 · out $0.4000 per 1M tokens | — |
whisper | ggml-large-v3-turbo-q5_0.bin | transcribe | 25,600 ms | 25,600 ms | 100 % (167 / 167) | not billed | — |
whisper | ggml-parakeet-tdt-0.6b-v3-q8_0.bin | transcribe | 25,600 ms | 25,600 ms | 100 % (989 / 989) | not billed | — |
whisper | whisper.cpp | stt | 25,600 ms | 25,600 ms | 0 % (0 / 51) | not billed | — |