Recipe
gemma-4-26b-a4b-it-q8-k-xl-intel-arc-pro-b70-32gb-llama-cpp-tp1
gemma-4-26b-a4b-it-q8-k-xl-intel-arc-pro-b70-32gb-llama-cpp-tp1Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.
Record
- Status
- candidate
- Source
- localmaxxing
- Engine
- llama.cpp
- Engine version
- c926ad098 local B70 SYCL/AOT Gemma patch stack
- Accelerators
- 1
- Tensor parallel
- 1
- Context tokens
- 32,768
- Max concurrency
- 1
- chat
- unknown
- reasoning
- unknown
- tools
- unknown
- vision
- unknown
Hugging Face model card
Identity
https://huggingface.co/unsloth/gemma-4-26B-A4B-it-GGUF- Repository
- unsloth/gemma-4-26B-A4B-it-GGUF
- 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
llama.cpp
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.
scripts/run-gemma4-26b-first-baseline.sh
Environment
| Variable | Value |
|---|---|
BATCH_SIZE | 1024 |
EXTRA_LLAMA_ARGS | --parallel 1 --cache-ram 0 --spec-type draft-mtp --spec-draft-model /mnt/fast-ai/llm-models/gemma4-26b-a4b-it-q8-gguf/MTP/gemma-4-26B-A4B-it-Q4_0-MTP.gguf --spec-draft-n-max 3 --spec-draft-device SYCL0 --spec-draft-ngl all --spec-draft-type-k f16 --spec-draft-type-v f16 --spec-draft-n-min 2 --spec-draft-p-min 0.0475 --no-spec-draft-backend-sampling --spec-draft-threads 32 --spec-draft-threads-batch 32 --ctx-checkpoints 0 |
GPU_INDEX | 0 |
LABEL | gemma4-q8-gpu0-finalpostnorm-reproexact-full512-20260701T084728Z |
LLAMA_SERVER | /home/steve/src/llama.cpp-gemma-record-repro-c926/build-sycl-b70-aot-bmg-g31-q8reorder-vdr2/bin/llama-server |
POLL | 100 |
PORT | 18530 |
REALISTIC_GATE | 1 |
REALISTIC_METRIC_TOKENS | 100 |
UBATCH_SIZE | 1024 |
Measured speed
| Concurrency | Context | Prefill | Decode | TTFT ms | Status | Sweep |
|---|---|---|---|---|---|---|
| 1 | 32,768 | — | 125 | 178.7 | observed | gemma-4-26b-a4b-it-q8-k-xl-intel-arc-pro-b70-32gb-llama-cpp-tp1-sweep |
Remaining fields
Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.
- hardware count
- 1
- hardware id
- intel-arc-pro-b70-32gb
- id
- gemma-4-26b-a4b-it-q8-k-xl-intel-arc-pro-b70-32gb-llama-cpp-tp1
- model instance id
- unsloth-gemma-4-26b-a4b-it-gguf--q8-k-xl
- recipe source
- localmaxxing
- schema version
- local-ai-registry/v1
- speed sweep ids
- gemma-4-26b-a4b-it-q8-k-xl-intel-arc-pro-b70-32gb-llama-cpp-tp1-sweep
- status
- candidate
capabilities
engine
- name
- llama.cpp
- version
- c926ad098 local B70 SYCL/AOT Gemma patch stack
serving
- max concurrency
- 1
- max context tokens
- 32,768
- tensor parallel
- 1
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
metadata
- localmaxxing · backend
- xpu
- localmaxxing · hardware label
- Intel Arc Pro B70
- localmaxxing · notes
- Gemma 4 26B A4B Q8/INT8-quality single-B70 realistic-suite result. Primary metric is median generated-token throughput for tokens 1-100 after TTFT across the fixed cold prompt suite. Each prompt was sent exactly once; every request reported cached_tokens=0; context checkpoints, prompt/KV cache reuse, response reuse, n-gram/history acceleration, and warmed repeated prompts were disabled. Target and verifier are UD-Q8_K_XL; only the MTP draft is Q4_0, and accepted tokens are verified by the target model. Supporting confirmations are listed in engineFlags.supportingConfirmationRuns.
- localmaxxing · observed command
- LLAMA_SERVER=/home/steve/src/llama.cpp-gemma-record-repro-c926/build-sycl-b70-aot-bmg-g31-q8reorder-vdr2/bin/llama-server GPU_INDEX=0 PORT=18530 LABEL=gemma4-q8-gpu0-finalpostnorm-reproexact-full512-20260701T084728Z BATCH_SIZE=1024 UBATCH_SIZE=1024 POLL=100 REALISTIC_GATE=1 REALISTIC_METRIC_TOKENS=100 EXTRA_LLAMA_ARGS='--parallel 1 --cache-ram 0 --spec-type draft-mtp --spec-draft-model /mnt/fast-ai/llm-models/gemma4-26b-a4b-it-q8-gguf/MTP/gemma-4-26B-A4B-it-Q4_0-MTP.gguf --spec-draft-n-max 3 --spec-draft-device SYCL0 --spec-draft-ngl all --spec-draft-type-k f16 --spec-draft-type-v f16 --spec-draft-n-min 2 --spec-draft-p-min 0.0475 --no-spec-draft-backend-sampling --spec-draft-threads 32 --spec-draft-threads-batch 32 --ctx-checkpoints 0' scripts/run-gemma4-26b-first-baseline.sh
- localmaxxing · run id
- cmr1u77na01k2ld01kalwzs1e
- localmaxxing · tokenized · arguments
- scripts/run-gemma4-26b-first-baseline.sh
- localmaxxing · tokenized · environment · BATCH SIZE
- 1024
- localmaxxing · tokenized · environment · EXTRA LLAMA ARGS
- --parallel 1 --cache-ram 0 --spec-type draft-mtp --spec-draft-model /mnt/fast-ai/llm-models/gemma4-26b-a4b-it-q8-gguf/MTP/gemma-4-26B-A4B-it-Q4_0-MTP.gguf --spec-draft-n-max 3 --spec-draft-device SYCL0 --spec-draft-ngl all --spec-draft-type-k f16 --spec-draft-type-v f16 --spec-draft-n-min 2 --spec-draft-p-min 0.0475 --no-spec-draft-backend-sampling --spec-draft-threads 32 --spec-draft-threads-batch 32 --ctx-checkpoints 0
- localmaxxing · tokenized · environment · GPU INDEX
- 0
- localmaxxing · tokenized · environment · LABEL
- gemma4-q8-gpu0-finalpostnorm-reproexact-full512-20260701T084728Z
- localmaxxing · tokenized · environment · LLAMA SERVER
- /home/steve/src/llama.cpp-gemma-record-repro-c926/build-sycl-b70-aot-bmg-g31-q8reorder-vdr2/bin/llama-server
- localmaxxing · tokenized · environment · POLL
- 100
- localmaxxing · tokenized · environment · PORT
- 18530
- localmaxxing · tokenized · environment · REALISTIC GATE
- 1
- localmaxxing · tokenized · environment · REALISTIC METRIC TOKENS
- 100
- localmaxxing · tokenized · environment · UBATCH SIZE
- 1024
- 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/cmr1u77na01k2ld01kalwzs1e ↗ |