Model result, rank 1 of 27

DeepSeek V4.1 Flash

OpenCode Go relay · Full + SWE + Hard measured. Public result card with the model’s overall score, lane measurements, runtime and cost telemetry, and the ranking formula.

Compare all models Download public JSON

Figure 1 Lane profile against the cohort Each bar is one measured lane on the shared 0 to 100 scale. The thin mark is the cohort median; the bright mark is the best score any ranked model reached on that lane.

Full / Agenticlane rank #2 of 27

89.83

SWE MVPlane rank #2 of 27

89.52

Hard Intelligencelane rank #10 of 27

88.11

DeepSeek V4.1 Flash
Cohort median
Cohort best

Overall score89.15

Rank #1

Full / Agentic89.83

Full rank #2

SWE MVP89.52

SWE rank #2

Hard Intelligence88.11

Hard rank #10

Measured cost$0.183

100.0% reliability

Overall

All-around publication view

Score89.15

Formulamean(Full, SWE, Hard Intelligence)

BasisOpenCode Go relay · Full + SWE + Hard measured

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

Lane 01

Full / Agentic benchmark

Final89.83

Capability98.52

Agentic95.51

Pass rate97.1%

Prompts43

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

Lane 02

Software engineering MVP

SWE score89.52

Focused final77.75

Capability88.88

Daily driver75.37

Prompts24

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

Lane 03

Hard Intelligence diagnostic

Hard score88.11

Active inquiry99.75

Online adaptation97.34

Self-repair85.34

Authority integrity70.00

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

Telemetry

Runtime economics

Total cost$0.183

Cost / scored item$0.0022

Seconds / timed item15.10s

Runtime coverage100.0%

Recorded tokens / item10.5k

Token coverage100.0%

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

Interpretation

Why the result lands here.

Rank #1 is the current all-around reference point: strong Full/Agentic performance, competitive SWE delivery, and transparent runtime telemetry. Hard Intelligence score is 88.11 and contributes to the overall score alongside Full/Agentic and SWE. Relay model: deepseek-v4.1-flash on the OpenCode Go relay. No effort level is published: the relay reports effectively flat reasoning-token counts across levels, so no effort label is claimed. Hard Intelligence is a public diagnostic, not a hidden official score.