Alibaba Qwen
QwQ 32B
Released Mar 6, 2025 · 31B tokens this week · 4 providers
Overview
A 32B model trained specifically for long, deliberate reasoning. It reaches well beyond its parameter count on mathematics and logic, at the cost of many thinking tokens. Best used with a generous output budget and a patient client.
Providers
Reference data — not yet measured from live traffic
| PROVIDER | MAX OUT | OUTPUT /M | QUANT | |||||
|---|---|---|---|---|---|---|---|---|
DeepInfraCheapest | 131K | 33K | $0.15 | $0.20 | 460ms | 120 | 98.70% | fp8 |
Hyperbolicdegraded | 66K | 33K | $0.16 | $0.21 | 640ms | 65 | 97.55% | fp4 |
Novita AI | 131K | 33K | $0.17 | $0.22 | 520ms | 100 | 98.30% | fp8 |
Groq | 33K | 33K | $0.18 | $0.24 | 209ms | 540 | 99.75% | Full |
Latency p50 reflects a provider’s API endpoint responsiveness — the round-trip to its API, not per-token inference time. These figures, with throughput and uptime, are reference data until the gateway aggregates its own traffic.
Price comparison
Blended $ per 1M tokens
Blended rate per 1M tokens, weighted one part prompt to three parts completion — roughly the shape of a chat workload. Your mix will move the number.
Call it
OpenAI-compatible — swap the base URL and go
curl https://model.cards/api/v1/chat/completions \
-H "Authorization: Bearer $MODELCARDS_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen/qwq-32b",
"messages": [
{ "role": "user", "content": "Summarize the tradeoffs of speculative decoding." }
]
}'