AL
AI21 Labs
Language model · Israel ·Mar 2024

Jamba

Language Open (Restricted)
51.6B
Parameters
8
Benchmarks

About Jamba

Jamba is an AI model developed by AI21 Labs, in the language category, released in 2024, made available as an open-weights model with 51.6B params.

On this page you'll find Jamba's full specifications, including 8 benchmark results. Review provider pricing and benchmark scores below, or compare Jamba head-to-head with other language models.

→ More models by AI21 Labs → All language models → AI benchmarks leaderboard

Links

Announcement ↗ Hugging Face ↗

Run Jamba locally — what to buy

Share this build ↗

Sized to what Jamba actually needs (~8k context at Q4), not to the biggest GPU: Good = cheapest that runs it, Better = best value, Best = most headroom. Speed is a hardware estimate; anything past ~120 tok/s is shown as instant.

Good
Apple M1 Max 64GB
Apple · ~$2,600
~11.4 tok/s
Q4 · fits full context · up to 8k ctx
Runs with: ExLlamaV2, KoboldCpp, LM Studio
See full estimate →
Better
Apple M2 Max 64GB
Apple · ~$2,800
~11.4 tok/s
Q4 · fits full context · up to 8k ctx
Runs with: ExLlamaV2, KoboldCpp, LM Studio
See full estimate →
Best
Apple M1 Ultra 64GB
Apple · ~$3,700
~22.7 tok/s
Q4 · fits full context · up to 8k ctx
Runs with: ExLlamaV2, KoboldCpp, LM Studio
See full estimate →

Software support

·ExLlamaV2 ·KoboldCpp ·LM Studio ·MLX ·Ollama ·SGLang ·TensorRT-LLM ·llama.cpp ·vLLM ?KTransformers ?text-generation-webui

✓ measured · · compatible · — not supported. Informational only — speed is hardware-based.

Consumer GPU
6.8–8.2 tok/s
2 options →
Apple
8.5–22.7 tok/s
15 options →
Datacenter GPU
23–120+ tok/s
13 options →

Community

Would you run Jamba again?
No votes yet — be the first.
Did it run? — community reports by hardware

No reports yet. Own this model on some hardware? Be the first to confirm it runs.

Discussion

No comments yet. Start the discussion.

Jamba benchmark scores▶
Compare on the benchmark table →
Benchmark Category Score Variant
HumanEval code 29.3% cited-qwen2-57b
MMLU general 67.4% cited-qwen2-57b
ARC-Challenge knowledge 64.4 cited-qwen2-57b
GSM8K math 59.9% cited-qwen2-57b
BIG-Bench Hard reasoning 45.4% cited-qwen2-57b
HellaSwag reasoning 87.1 cited-qwen2-57b
Winogrande reasoning 82.5 cited-qwen2-57b
TruthfulQA safety 46.4% cited-qwen2-57b

More models like Jamba

AI
DMind-1
 AI
COLLECTIVE-1
 AI
Cogito 1 series
 AI
Euler (Euler大模型)
· 2.5B
 AI
SmolLM-1.7B
· 1.7B
 AI
Llama-3-70B-Synthia-v3.5
· 70B

Frequently asked questions

What is Jamba? ▶

Jamba is an AI model developed by AI21 Labs, in the language category, released in 2024. It is tracked on GenAIList with its specifications, benchmark scores and provider pricing.

Who created Jamba? ▶

Jamba was developed by AI21 Labs and released in 2024.

Is Jamba open source or proprietary? ▶

Jamba ships as an open-weights model: you can download and self-host the weights, though the licence may place some restrictions on use.

How much does Jamba cost? ▶

Pricing for Jamba depends on the provider. See the providers table on this page for the latest API rates.

How does Jamba perform on benchmarks? ▶

Jamba is benchmarked across 8 evaluations on GenAIList, including HumanEval (29.3%). See the full benchmark table below and compare it with other models.

No videos yet. Know a good one? Add it below.

Know a great video about Jamba?