Model result · rank #15

Ornith‑1.0‑35B Q4_K_M

Local model

Local GGUF · llama.cpp Vulkan · Q4_K_M · 35B MoE. Public result card with the model’s overall score, lane measurements, runtime/cost telemetry, and ranking formula.

Download public JSON Compare all models

Overall score55.57

Rank #15

Full / Agentic77.96

Full rank #14

SWE MVP39.16

SWE rank #16

Hard Intelligence49.58

Hard rank #12

Measured cost$0

100.0% reliability

Overall

All-around publication view

Score55.57

Formulamean(Full, SWE, Hard Intelligence)

BasisLocal GGUF · llama.cpp Vulkan · Q4_K_M · 35B MoE

The overall score averages the measured major lanes while keeping each source measurement visible.

Lane 01

Full / Agentic benchmark

Final77.96

Capability88.34

Agentic79.66

Pass rate79.1%

Prompts43

This lane captures instruction following, structured behavior, tool discipline, and general agentic reliability.

Lane 02

Software engineering MVP

SWE score39.16

Focused final39.16

Capability17.79

Daily driver45.49

Prompts24

This lane is closer to implementation usefulness: source handling, architecture cleanliness, and deliverable quality.

Lane 03

Hard Intelligence diagnostic

Hard score49.58

Active inquiry18.75

Online adaptation11.25

Self-repair78.33

Authority integrity90.00

Hard Intelligence measures active inquiry, online adaptation, evidence-driven self-repair, and authority/salience integrity.

Telemetry

Runtime economics

Total cost$0

Cost / scored item$0

Seconds / timed item117.66s

Runtime coverage100.0%

Recorded tokens / item8.7k

Token coverage100.0%

Cost, time, and token basis are normalized telemetry. They explain tradeoffs; they do not overwrite the capability score yet.

Interpretation

Why this result lands here.

The model is stronger in the Full/Agentic lane than in the SWE lane; the overall score is therefore shown with both component lanes visible. Hard Intelligence score is 49.58 and contributes to the overall score alongside Full/Agentic and SWE. Local model row: benchmarked on local hardware with no API metering. Strong local Full/Agentic baseline for a 35B MoE GGUF, but SWE implementation/review and Hard Intelligence inquiry/adaptation were weak in this run. The local entrant shares the unified public tournament table with API-backed entrants while exposing local runtime and cost metadata.