The Open-Source AI Stack
RSS
All models

Models · Compare

Claude Haiku 4.5 vs Granite 4.0 Small

Rows highlighted in warm gray are where the models differ. Numbers carry their as-of date and primary source.

Specs

Field A: Claude Haiku 4.5 B: Granite 4.0 Small
Released 2025-10-152025-10-02
Developer AnthropicIBM
Openness ProprietaryOpen
License ProprietaryApache-2.0
OSI-approved noyes
Data released nono
Training code nono
Architecture unknownhybrid-mamba-transformer-moe
Total params 32B
Active params 9B
Experts
Context window 200K131K
Attention unknownhybrid-mamba2-transformer
Position enc. unknownunknown
Pretraining tokens
Post-training rlhf, constitutionalsft, rlhf
Training hardware
$/M input $1.00
$/M output $5.00
Output tok/sec

Benchmarks

Missing scores render as not reported; never inferred. Bold highlights the leader per benchmark.

Code

SWE-Bench Verified 73.3 2025-10-15

Context · A

Small-tier 4.5 model released October 15 2025 at $1 / $5 per million tokens, one-third the Sonnet 4 price. Anthropic positioned it as matching Sonnet 4 on coding and computer use while running 4-5x faster than Sonnet 4.5. First Haiku to include extended thinking, computer use, and explicit context awareness.

Context · B

IBM's first hybrid Mamba/Transformer release, October 29 2025. Granite-4.0-H-Small is a 32B MoE activating 9B per token, with Mamba-2 and conventional transformer blocks layered 9:1. Shipped alongside H-Tiny (7B / 1B) and H-Micro (3B dense), all base + instruct, all Apache 2.0.

Claude Haiku 4.5 detail → · Granite 4.0 Small detail →