d1-3B on local hardware
Estimated comfort for repository coding with Aider on a MacBook Pro M4 Pro 48GB.
d1-3B
LiquidAI · other · 22,004 downloads
VisionReasoning / thinking mode
- Parameters
- 3.1B exact
- Context window
- 32K tokens
- Attention
- Standard (full attention)
- Layers · KV heads · head dim
- 30 · 8 · 64
- KV cache
- 16 KB/token · 0.5 GB at 32K
- Tool calling
- good (from chat template)
- Thinking
- good (from chat template)
- Vision
- good (from config)
- Commercial use
- check cardCheck the model card
- Import confidence
- medium · capabilities estimated
Weights by quantization
Q4_K_M
1.6 GB
gguf file size
Q8_0
2.7 GB
gguf file size
F16 / BF16
5.0 GB
gguf file size
Q3_K_M
1.3 GB
gguf file size
Q5_K_M
1.8 GB
gguf file size
Q6_K
2.1 GB
gguf file size
- • vision: good, inferred from config (low confidence).
- • Capability ratings are estimated from model size and name — this model hasn't been assessed by hand.
- • Commercial use unknown: Check the model card.
- • GGUF sizes come from community conversions matched by model name; equivalence and runtime support are not verified. MLX conversion availability is unverified.
d1-3B — 8-bit
Borderline for your workload
Speed
45–65 tok/s
Excellent
Memory left
32 GB
of 48 GB
First word in
≈15 s
Comfortable
All the details
- Your setup
- MacBook Pro M4 Pro 48 GB
- Workload
- Repository coding with Aider
- Runtime → tool
- LM Studio → Aider
- Memory used
- 4.7 GB
- Context
- 32K · up to 32K practical
- Tool support
- Official
- Use-case fit
- Poorly suited
Why?
- Model fits with approximately 32 GB of memory headroom.
- LM Studio supports hardware acceleration on this machine (Metal (Apple GPU)).
- Aider can connect to LM Studio (officially supported).
- Generation (45–65 tok/s) should be appropriate for interactive coding.
- Ingesting a prompt of 14K tokens takes ≈15 s; prompt processing keeps up with interactive coding.
- 32K working context is reasonable.
- Weak coding capability for interactive coding.
Why “Borderline” rather than “Acceptable”?
- Workload suitability: Weak coding capability for interactive coding.
How sure are we?
Medium confidenceEstimated from bandwidth and compute, calibrated against 2 verified benchmarks of other models on this same chip (measured with a different engine, so calibration is partial).
- Calibrated with verified benchmarks of other models on this chip.
- Hybrid/sliding attention performance varies more between runtimes.
Medium confidence