Benchmark
Submitted bybasecompute
Public report
This is the public version of the benchmark. It includes model, runtime, protocol, and throughput information, without telemetry, raw samples, or signing metadata.
Compare throughput at the same token counts. A signed report does not independently verify benchmark execution.
Model
- Canonical ID
- meta-llama/Llama-3.2-3B-Instruct
- Identity resolution
- artifact_verified
- Identity verification
- verified
- Verified at
- Name
- Llama-3.2-3B-Instruct-Q8
- Architecture
- llama
- Quantization
- base_q4
- Artifact
- Provider
- huggingface
- Repo ID
- basecompute/Llama-3.2-3B-Instruct
- Revision
- c111e66fb8f31b78e374d67d8145504998787ea6
- Path
- Llama-3.2-3B-Instruct-Q8.base
- SHA256
- b26f0dd86648cec9d1f7f30dbf986c59077395dffe8607a28640a2e8d251177a
- Format
- base
- Quantization namespace
- basert
- Quantization scheme
- base_q4
Runtime
- Name
- basert
- Version
- 0.2.6
Benchmark
- Mode
- text
- Chip
- Apple M1 Max
- Backend
- metal
- Protocol
- Report schema
- computearena-benchmark/1
- Benchmark schema
- basert-benchmark-harness/1
- Telemetry schema
- basert-telemetry/4
- Throughput schema
- computearena-throughput/3
- Comparable
- Yes (true)
- Context isolation
- headline_then_per_prefill
- Execution layout
Report metadata
- Schema
- computearena-public-benchmark/1
- Submission ID
- fe9b25b8-b82b-45f8-9e3b-bd803d5c1e5d
- Run ID
- 7907ad1c126dc70007c1775f7505bb1e
- Benchmark date
- (Unix ms: 1789875664533)
- Submitted by
- prabod