deepreinforce-ai/ornith-1.0-35b
ornith-1.0-35b (llamacpp)
Throughput
50.1t/s@ 32k
Engine
llama.cpp
Parameters
35B
Released
2026-06-21
Benchmarked
2026-06-27
Context ladder
Throughput at each benched context window.
| Context | KV | Throughput |
|---|---|---|
| @ 32k peak golden | q8_0 | 50.1t/s |
| @ 64k | q8_0 | 16.6t/s |
| @ 96k | q8_0 | 11.1t/s |
| @ 128k | q8_0 | 7.8t/s |
| @ 192k | q8_0 | 5.4t/s |
| @ 256k | q8_0 | 3.9t/s |
Golden profile
deepreinforce-ai-ornith-1-0-35b-llama
Capabilities
labllamacppgoldengguf
Why we run it
Golden fleet target — auto-scaffolded from recipe deepreinforce-ai-ornith-1-0-35b-llama.
Bench notes
golden 32k/q8_0 @ 50.1 tok/s — bench-agent-v2