REAL LOCAL SUITES · CODING V2 + COMPREHENSIVE · SUITES NOT MIXED

BENCHMARK PLANNER

Plan the work. Execute no queue.

The persistent queue remains planning-only. Phase 2B.1 can run one explicitly named canary, but cannot start, advance, or batch this queue.

QueueHBQ-UNINITIALIZED0 planned items
ModePLANNING ONLYexecutionEnabled = false
Eligible models15available / not benchmarked
Concurrency policy1future preview policy

PERSISTED QUEUE

Queue is empty

Execution disabled

No models are queued

Use the planning CLI to add selected registry models. Saving a plan cannot load or benchmark them.

heavrybench plan <model>
Resume metadataNext pending: noneInterrupted item: noneStop reason: none

REGISTRY CANDIDATES

Available to plan

15 models

FULL BENCHMARK QUEUE · NOT ACTIVE

Fail closed lifecycle

STOP QUEUE on failure
01DISCOVER
02PLAN
03PREFLIGHT
04LOAD ONE MODEL
05WARMUP
06BENCHMARK
07SAVE ATOMICALLY
08REQUEST UNLOAD
09VERIFY RELEASE WITH API PS
10NEXT MODEL
Default stop conditionsoomcrashtimeoutunload failureunexpected api ps stateresource release failure