Model Details
Amazon Nova 2 Pro
Quick answer: Amazon Nova 2 Pro is Amazon's June 2025 pro-tier model scoring 89.0% HumanEval, 74.6% LiveCodeBench, 81.6% MMLU-Pro, 80.2% IFBench, and 41.3% TerminalBench. Available via Amazon Bedrock.
At a Glance
Where Nova 2 Pro leads
- 89.0% HumanEval — strong Python coding
- 74.6% LiveCodeBench — competitive real-world coding
- 81.6% MMLU-Pro — good academic breadth
- 80.2% IFBench — solid instruction following
- Tight AWS/Bedrock integration
Where it lags
- 4.4% HLE — very low frontier reasoning
- 41.3% TerminalBench — limited agentic terminal tasks
- Amazon Bedrock only: no open weights
Best for: AWS-native coding and instruction-following pipelines; Bedrock users needing high HumanEval with good MMLU-Pro coverage.
What Amazon Nova 2 Pro Is
Amazon Nova 2 Pro is the "Pro" tier of Amazon's Nova 2 generation, released June 2025. It sits above Nova Lite/Micro in Amazon's model family. The 89.0% HumanEval and 74.6% LiveCodeBench demonstrate strong coding ability within the AWS ecosystem.
The 4.4% HLE score reflects the model's focus on instruction/coding rather than frontier reasoning research tasks.
Specifications
| Field | Value |
|---|---|
| Organization | Amazon |
| License | Amazon Bedrock (Proprietary) |
| Release date | June 2025 |
| Modality | Text only |
Pricing
Available via Amazon Bedrock. Check current pricing at aws.amazon.com/bedrock.
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| HumanEval | 89.0% | Benchgen evaluation | 2025-06 |
| LiveCodeBench | 74.6% | Benchgen evaluation | 2025-06 |
| MMLU-Pro | 81.6% | Benchgen evaluation | 2025-06 |
| IFBench | 80.2% | Benchgen evaluation | 2025-06 |
| TerminalBench | 41.3% | Benchgen evaluation | 2025-06 |
| Humanity's Last Exam | 4.4% | Benchgen evaluation | 2025-06 |
Nova 2 Pro vs Alternatives
| Model | HumanEval | LiveCodeBench | MMLU-Pro | Access |
|---|---|---|---|---|
| Nova 2 Pro | 89.0% | 74.6% | 81.6% | Amazon Bedrock |
| MiniMax M2 | — | 83.0% | 82% | MiniMax API |
| Qwen2.5 Coder 32B Instruct | 92.7% | — | — | Open / API |
Nova 2 Pro vs MiniMax M2: Nova 2 Pro has higher HumanEval (89.0%) but lower LiveCodeBench (74.6% vs 83.0%). For AWS workloads: Nova 2 Pro. For raw LiveCodeBench: MiniMax M2.
Related Models & Benchmarks
- Compare: MiniMax M2 · Qwen2.5 Coder 32B Instruct
- Benchmarks: HumanEval · LiveCodeBench · IFBench
Frequently Asked Questions
What is Amazon Nova 2 Pro?
Amazon's June 2025 Bedrock model scoring 89.0% HumanEval, 74.6% LiveCodeBench, 81.6% MMLU-Pro. AWS-native coding and instruction-following.Specs from Amazon's Nova 2 Pro release (June 2025) and Benchgen evaluations. Last updated 2026-07-24.
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.