Open-Source Alternatives to 
Claude Haiku 4.5
The best open-weight, self-hostable alternatives to Claude Haiku 4.5 in 2026 — compared on price, context window and capabilities. Run them locally and cut API costs.
Refreshed from live data · olud.ai
The best open-weight alternatives to Claude Haiku 4.5
Claude Haiku 4.5 is a proprietary, API-only model. These open-weight models can be self-hosted, run offline and used at a fraction of the cost — here's how the top ones stack up.
DeepSeek V4.1 Flash Open
21.9 pts ABOVE Claude Haiku 4.5 on the Artificial Analysis intelligence index · 8.3× cheaper per million output tokens
DeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. It activates 8B parameters on input and 16B on...
DeepSeek V4.1 Flash vs Claude Haiku 4.5 →
MiniMax M3 Open
12.0 pts ABOVE Claude Haiku 4.5 on the Artificial Analysis intelligence index · 4.2× cheaper per million output tokens
MiniMax-M3 is a multimodal foundation model from MiniMax. It supports text, image, and video inputs with text output, a 1M-token context window, and is suited for long-horizon agentic work, coding,...
MiniMax M3 vs Claude Haiku 4.5 →TH Inkling Small Open
8.5 pts ABOVE Claude Haiku 4.5 on the Artificial Analysis intelligence index · 4.2× cheaper per million output tokens
Inkling Small is an open-weight multimodal mixture-of-experts model from Thinking Machines Lab, with 12B active parameters out of 276B total. It is positioned as the smaller, more efficient member of...
Inkling Small vs Claude Haiku 4.5 →
Step 3.7 Flash Open
1.9 pts ABOVE Claude Haiku 4.5 on the Artificial Analysis intelligence index · 4.3× cheaper per million output tokens
Step 3.7 Flash is StepFun's latest high-efficiency multimodal Mixture-of-Experts model. It pairs a 196B-parameter language backbone with a vision encoder for native image and video understanding, activating roughly 11B parameters...
Step 3.7 Flash vs Claude Haiku 4.5 →
Gemma 4 26B A4B Open
Within 0.9 pts of Claude Haiku 4.5 on the Artificial Analysis intelligence index · 22.7× cheaper per million output tokens
Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind. Despite 25.2B total parameters, only 3.8B activate per token during inference — delivering near-31B quality at...
Gemma 4 26B A4B vs Claude Haiku 4.5 →See the full open-source leaderboard
Live ranking of open-weight models with pricing, context windows and capabilities.
Open the leaderboard →