AI21-Jamba-Reasoning-3B
Where it runs 8K context · ≥ 5 t/s floor
| Setup | Best plan | Memory | Generate |
|---|---|---|---|
| 8 GB GPU + 32 GB RAM | FP16llama.cpp CUDAFull GPU | KV unknown | ~26–37 t/s Estimated |
| 12 GB GPU + 32 GB RAM | FP16llama.cpp CUDAFull GPU | KV unknown | ~43–46 t/s Estimated |
| 16 GB GPU + 64 GB RAM | FP16llama.cpp CUDAFull GPU | KV unknown | ~89–109 t/s Estimated |
| 24 GB GPU + 64 GB RAM | FP16llama.cpp CUDAFull GPU | KV unknown | ~90–113 t/s Estimated |
| 32 GB GPU + 128 GB RAM | FP16llama.cpp CUDAFull GPU | KV unknown | ~141–226 t/s Estimated |
| 2× 24 GB GPUs + 128 GB RAM | FP16llama.cpp CUDAFull GPU | KV unknown | ~82–106 t/s Estimated |
| 96 GB workstation GPU + 256 GB RAM | FP16llama.cpp CUDAFull GPU | KV unknown | ~145–227 t/s Estimated |
| Mac, 24 GB | FP16llama.cpp MetalUnified memory | KV unknown | ~30–36 t/s Estimated |
| Mac, 48 GB | FP16llama.cpp MetalUnified memory | KV unknown | ~30–36 t/s Estimated |
| Mac, 128 GB | FP16llama.cpp MetalUnified memory | KV unknown | ~49–75 t/s Estimated |
| Mac, 512 GB | FP16llama.cpp MetalUnified memory | KV unknown | ~55–84 t/s Estimated |
| DGX Spark 128 GB | FP16llama.cpp CUDAUnified memory | KV unknown | ~32–36 t/s Estimated |
| Ryzen AI Max+ 395, 128 GB | FP16llama.cpp VulkanUnified memory | KV unknown | ~22–32 t/s Estimated |
| CPU only, 64 GB DDR5-5600 dual channel | Q4_K_Mllama.cpp CPUCPU only | KV unknown | ~11–12 t/s Estimated extrapolated |
Same plan engine as the planner, with its default reserves and RAM. How plans are chosen.
Files exact Hugging Face bytes
| Artifact | Size | Files | Revision | Source |
|---|---|---|---|---|
| FP16 ai21labs/AI21-Jamba-Reasoning-3B-GGUF | 5.96 GiB | 1 | 462e08a43c | HF files |
| Q4_K_M ai21labs/AI21-Jamba-Reasoning-3B-GGUF | 1.80 GiB | 1 | 462e08a43c | HF files |
| Native safetensors BF16 ai21labs/AI21-Jamba-Reasoning-3B | 5.96 GiB | 2 | 7524370414 | HF files |
Not found in the representative quant search (not proven unavailable): GGUF Q8_0, GGUF Q5_K_M, MLX 8-bit, MLX 4-bit, EXL3, AWQ, GPTQ. Complete selected file sets; projectors, MTP/draft and lookup files are listed separately and budgeted only by plans that use them. File bytes are not allocated memory.
Quality
| AA evaluated mode | Index | LiveCodeBench | Terminal-Bench 2.1 | GPQA | HLE | Source |
|---|---|---|---|---|---|---|
| Jamba Reasoning 3B AA estimate | 5.7 | 6.3% | — | 33.3% | 3.8% | source |
Arena text rating
Unknown
Publisher-reported benchmarks
None extracted from the model card.
AA and Arena describe the hosted model in the stated mode, not each quantized file; scores are never inherited by base, pruned, abliterated or quantized variants and never averaged together.
How to run it
No model-specific recipe in the research. The planner binds the backend’s documented command template (llama.cpp, Ollama, LM Studio, MLX, vLLM) to your plan: open a plan’s details.
Recipes are source examples (settings are the publisher’s or author’s, not measured optima) and were not executed by LocalFit. A downloadable file does not prove your backend build supports the architecture.
Reported speeds 0
No reported measurement yet. Submit one.