zai-org/GLM-5.2

Text Generation • 753B • Updated Sep 1 • 682k • • 5.15k

Note $1.00 in / $3.00 out per 1M. The step up when an 8B to 35B is not enough. Tool calling and structured output verified against our API.

deepseek-ai/DeepSeek-V4-Pro

Text Generation • 1.6T • Updated Jun 22 • 436k • • 5.64k

Note $0.91 in / $2.72 out per 1M. Tool calling and structured output verified against our API.

Qwen/Qwen3.6-35B-A3B

Image-Text-to-Text • 36B • Updated Apr 24 • 3.52M • • 2.94k

Note 35B MoE, 3B active. 170 ms TTFT p95, 4 ms per token, ~213 tok/s, FP8. $0.40 in and out per 1M. Larger and faster than our 8B dense model - sparse MoE means parameter count stopped predicting speed.

Qwen/Qwen3-Coder-30B-A3B-Instruct

Text Generation • 31B • Updated Dec 3, 2025 • 412k • • 1.27k

Note 30B MoE, 3B active. 30 ms TTFT p95 - the lowest in our catalogue - 9 ms per token, ~111 tok/s, BF16. $0.10 in / $0.30 out per 1M, 32k context. TTFT is what agent loops pay on every step of a plan.

openai/gpt-oss-20b

Text Generation • 21B • Updated Aug 26, 2025 • 6.19M • • 5.14k

Note 20B MoE, 3.6B active. 170 ms TTFT p95, 8 ms per token, ~125 tok/s, MXFP4 as the vendor ships it. $0.20 in / $0.28 out per 1M, 128k context. Peak 8,900 tok/s under concurrency on one card.

Qwen/Qwen3-VL-8B-Instruct

Image-Text-to-Text • 9B • Updated Oct 15, 2025 • 10.4M • • 1.17k

Note 8B vision-language. 162 ms TTFT p95, 12 ms per token, ~78 tok/s, BF16. $0.15 in / $0.50 out per 1M. Documents, charts, screenshots and photographs.

Note Live demo. Streaming chat with image input, reporting time to first token, throughput, token counts and the exact cost of each reply.