vs
1:0
overall capability wins (2 ties across 3 scored). Jamba Large 1.7 leads overall.
- Pricing0:3
- Speed0:2
- Specs1:0
Verdict
Jamba Mini 2 is ~14.0x cheaper on a blended token basis than Jamba Large 1.7.
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Jamba Large 1.7
- $4
- Jamba Mini 2
- $0.30
Jamba Mini 2 is about 13.3x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
AI21
AI21
Family
Jamba
Jamba
License
Jamba Open Model
Jamba Open Model
Open weights
Yes
Yes
Release
Jul 8, 2025
Jan 1, 2026
Knowledge cutoff
-
-
API / provider
AI21
AI21
Modalities
text → text
text → text
Specs
Context window
256K
256K
Max output
4K
4K
Parameters
398B (94B act.)
52B (12B act.)
Pricing
Input $/1M
$2
$0.20
Output $/1M
$8
$0.40
Blended $/1M (3∶1)
$3.50
$0.25
Speed
tok/s
57
100
TTFT (s)
0.87
0.35
Reasoning
MMLU-Pro
—
—
GPQA Diamond
—
—
Humanity's Last Exam
—
—
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
—
—
AA-LCR v1.1
—
—
CritPt
—
—
MMMU-Pro
—
—
IFBench
—
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
—
—
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
—
—
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
—
| Metric | Jamba Large 1.7 | Jamba Mini 2 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | AI21 | AI21 | — |
| Family | Jamba | Jamba | — |
| License | Jamba Open Model | Jamba Open Model | — |
| Open weights | Yes | Yes | — |
| Release | Jul 8, 2025 | Jan 1, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | AI21 | AI21 | — |
| Modalities | text → text | text → text | — |
| Specs | |||
| Context window | 256K | 256K | tie |
| Max output | 4K | 4K | tie |
| Parameters | 398B (94B act.) | 52B (12B act.) | A +346 |
| Pricing | |||
| Input $/1M | $2 | $0.20 | B +$1.80 |
| Output $/1M | $8 | $0.40 | B +$7.60 |
| Blended $/1M (3∶1) | $3.50 | $0.25 | B +$3.25 |
| Speed | |||
| tok/s | 57 | 100 | B +43 |
| TTFT (s) | 0.87 | 0.35 | B +0.52 |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | — | — | — |
| Humanity's Last Exam | — | — | — |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | — | — | — |
| AA-LCR v1.1 | — | — | — |
| CritPt | — | — | — |
| MMMU-Pro | — | — | — |
| IFBench | — | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | — | — | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | — | — | — |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | — | — |
Jamba Large 1.7
- Context
- 256K / 4K out
- Parameters
- 398B (94B act.)
- Price
- $2 / $8
- Speed
- 57 tok/s · 0.87s TTFT
- Modalities
- text → text
- License
- Jamba Open Model
AI21's flagship hybrid Mamba-Transformer MoE — 256K context with efficient long-document inference versus pure attention stacks.
Jamba Mini 2
- Context
- 256K / 4K out
- Parameters
- 52B (12B act.)
- Price
- $0.20 / $0.40
- Speed
- 100 tok/s · 0.35s TTFT
- Modalities
- text → text
- License
- Jamba Open Model
Compact Jamba 2 MoE with full 256K window — one of the best context-per-dollar ratios for enterprise doc workflows.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.
These two models have no overlapping published benchmarks in our dataset. Compare specs and pricing instead.