Providers Parasail (fp8) via OpenRouter
2 models measured.
qwen/qwen3.6-35b-a3b
- Images
- 170 “This endpoint's maximum context length is 262144 tokens. However, you requested about 263079 tokens (407 of text input, 262656 of image input, 16 in the output). Please reduce the length of either one, or use the context-compression plugin to compress your prompt automatically.”
- 768×432 frames
- —
- Tokens / frame
- —
- Inline body
- —
- Images by URL
- —
- Video file
- —
- Video tok/min
- Rejected “Request decoded media exceeds VLLM_MAX_MEDIA_REQUEST_DECODE_BYTES=715827880 (3832012800 bytes).”
- Frames/min
- —
- Video input
- Unknown
- Tool messages
- —
- Extra fields
- All accepted
Measured 2026-10-31
qwen/qwen3.8-27b
- Images
- 170 “No endpoints found for qwen/qwen3.8-27b. Every candidate endpoint was removed during routing: Filter by Context Length removed darkbloom/fp4, chutes/fp8, near-ai/fp8, ionstream/fp8, reka/fp8, modal, dekallm, wafer/fp8, deepinfra/bf16, tenstorrent/fp4, mancer/fp8, modelrun/fp4, akashml/fp8, parasail/fp8, coreweave/fp8, cloudflare, venice/fp8, cerebras/fp16; Filter by Fallback removed novita, alibab”
- 768×432 frames
- —
- Tokens / frame
- —
- Inline body
- —
- Images by URL
- —
- Video file
- —
- Video tok/min
- 2,499
- Frames/min
- 6.3 ⁎
- Video input
- Yes
- Tool messages
- —
- Extra fields
- All accepted
Measured 2026-10-31