Recipe
qwen3-6-35b-a3b-mq4r-radeon-ai-pro-r9700-32gb-hipfire-tp4
qwen3-6-35b-a3b-mq4r-radeon-ai-pro-r9700-32gb-hipfire-tp4Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.
Record
- Status
- candidate
- Source
- localmaxxing
- Engine
- hipfire
- Engine version
- 0.3.0+b0bcc3f91506
- Accelerators
- 4
- Tensor parallel
- 4
- chat
- unknown
- reasoning
- unknown
- tools
- unknown
- vision
- unknown
Hugging Face model card
Identity
https://huggingface.co/Qwen/Qwen3.6-35B-A3B- Repository
- Qwen/Qwen3.6-35B-A3B
- Status
- known
- Link type
- Exact Hub repository
Public Hugging Face repository confirmed by the Hub API.
Candidate: useful compatibility or speed evidence. The registry does not offer Run until promotion requirements are met.
Observed configuration
hipfire
Evidence only · candidate · reference
Candidate evidence — not a Run contract
Observed source tokens
Mechanical split of the source command. Unverified against the engine CLI. The registry does not offer Run for this recipe.
hipfire-batch-generate--model~/.hipfire/models/qwen3.6-35b-a3b.mq4r--batch112--max-seq1024--max-new64--devices0,1,2,3--data-parallel--kvq8--fresh-process-per-wave
| Flag | Value |
|---|---|
--model | ~/.hipfire/models/qwen3.6-35b-a3b.mq4r |
--batch | 112 |
--max-seq | 1024 |
--max-new | 64 |
--devices | 0,1,2,3 |
--kv | q8 |
Measured speed
| Concurrency | Context | Prefill | Decode | TTFT ms | Status | Sweep |
|---|---|---|---|---|---|---|
| 112 | 1,024 | — | 53.6 | 120 | observed | qwen3-6-35b-a3b-mq4r-radeon-ai-pro-r9700-32gb-hipfire-tp4-sweep |
Remaining fields
Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.
- hardware count
- 4
- hardware id
- radeon-ai-pro-r9700-32gb
- id
- qwen3-6-35b-a3b-mq4r-radeon-ai-pro-r9700-32gb-hipfire-tp4
- model instance id
- qwen-qwen3-6-35b-a3b--mq4r
- recipe source
- localmaxxing
- schema version
- local-ai-registry/v1
- speed sweep ids
- qwen3-6-35b-a3b-mq4r-radeon-ai-pro-r9700-32gb-hipfire-tp4-sweep
- status
- candidate
capabilities
engine
- name
- hipfire
- version
- 0.3.0+b0bcc3f91506
serving
- tensor parallel
- 4
Provenance & metadata (3)
facts
- capabilities.chat · provenance · captured at
- 2026-08-30T09:10:02Z
capabilities.chat · reason capability-not-verified
capabilities.chat · state unknown
- capabilities.reasoning · provenance · captured at
- 2026-08-30T09:10:02Z
capabilities.reasoning · reason capability-not-verified
capabilities.reasoning · state unknown
- capabilities.tools · provenance · captured at
- 2026-08-30T09:10:02Z
capabilities.tools · reason capability-not-verified
capabilities.tools · state unknown
- capabilities.vision · provenance · captured at
- 2026-08-30T09:10:02Z
capabilities.vision · reason capability-not-verified
capabilities.vision · state unknown
- engine.graph mode · provenance · captured at
- 2026-08-30T09:10:02Z
engine.graph mode · reason runtime-detail-not-published
engine.graph mode · state unknown
- serving.kv cache tokens · provenance · captured at
- 2026-08-30T09:10:02Z
serving.kv cache tokens · reason kv-cache-capacity-not-published
serving.kv cache tokens · state unknown
- serving.max concurrency · provenance · captured at
- 2026-09-01T10:20:39Z
serving.max concurrency · reason server-capacity-not-evidenced
serving.max concurrency · state unknown
- serving.max context tokens · provenance · captured at
- 2026-09-01T10:20:39Z
serving.max context tokens · reason context-limit-not-evidenced
serving.max context tokens · state unknown
metadata
- localmaxxing · backend
- rocm
- localmaxxing · batch size
- 112
- localmaxxing · hardware label
- Radeon AI Pro R9700
- localmaxxing · notes
- BATCHED AGGREGATE THROUGHPUT — not single-stream decode. DATA-PARALLEL across 4 cards; aggregate of 4 independent replicas. batch=112, concurrency=112, 100 input rows, 3 waves, max_tokens=64, ctx=1024, KV=q8, greedy; TTFT not comparable. Per-card median MODEL tok/s: d0:1495.3, d1:1504.4, d2:1500.1, d3:1501.6 (sum 6001.4); wall-clock aggregate 3889.4 tok/s (per-card wall d0:969.6, d1:975.7, d2:973.2, d3:970.9). MODEL tok/s is sum of per-card decode rates excluding inter-wave gap; wall includes all overhead. Do not compare DP aggregate to single-stream. engine: hipfire 0.3.0+b0bcc3f91506, backend rocm, gfx1201 x4 quant: MQ4R = MagnumQuant 4-bit Redline (graded mixed-tier: MQ4 attn/router/shared + graded MQ4 routed experts, Redline retained-PM4 dispatch), 4.16 bpw effective (18700048128 bytes *8 / 35.95B params) model file: qwen3.6-35b-a3b.mq4r sha256 4685c140c46b1a6f prompt: uniform short-serve jsonl md5 3d22208dcf539818ff2e8341f53474ee, 64 max_new method: median of 3 waves, fresh process per wave, temperature 0, bit-exact outputs (sha ec47dcfb9b0c54e0) evidence: sealed case a3b-mq4r-dp4-b112-gfx1201-r9700-4x, corpus manifest manifest-sha12-8f3c2e1a4b9d not measured: peak VRAM (not instrumented this campaign)
- localmaxxing · observed command
- hipfire-batch-generate --model ~/.hipfire/models/qwen3.6-35b-a3b.mq4r --batch 112 --max-seq 1024 --max-new 64 --devices 0,1,2,3 --data-parallel --kv q8 --fresh-process-per-wave
- localmaxxing · run id
- cmsnp1rko00jco001mvymwmsb
- localmaxxing · tokenized · arguments
- hipfire-batch-generate, --model, ~/.hipfire/models/qwen3.6-35b-a3b.mq4r, --batch, 112, --max-seq, 1024, --max-new, 64, --devices, 0,1,2,3, --data-parallel, --kv, q8, --fresh-process-per-wave
- localmaxxing · tokenized · fidelity
- faithful
provenance
- captured at
- 2026-08-30T09:10:02Z
sources
| captured at | kind | url |
|---|---|---|
| 2026-08-30T09:10:02Z | normalized-recipe | www.localmaxxing.com/en/runs/cmsnp1rko00jco001mvymwmsb ↗ |