Falcon H1 34B Instruct
Technology Innovation Institute
Key facts · verified Aug 15, 2026
Falcon H1 34B Instruct is a Technology Innovation Institute model released on 21 May 2025. Pricing: Open weights, self-hosted. Its context window is 256k tokens. It is not yet ranked on LMArena's text leaderboard. Source: huggingface.co.
TII's Falcon H1, a family of hybrid Transformer-Mamba (SSM) language models; the 34B Instruct variant is the flagship, with a 256K-token context window. TII's own model card reports 87.2% on HumanEval, 83.86% on MBPP and 49.71% on LiveCodeBench, and states the hybrid architecture gives up to 4x input and 8x output throughput versus a similarly-sized Transformer (Qwen2.5-32B). Text-only, no vision input.
Pricing & specs checkedSource: vendor documentation · checked Aug 15, 2026
Pricing
Open weights, self-hosted
Context
256k tokens
LMArena Elo
Not rated
LMArena rank
Not ranked
Key Features
Open Weights (TII Falcon License) Hybrid Transformer-Mamba 256K Context Multilingual (18 Languages) Self-Hostable High-Throughput Inference
Full Specifications
Open weights, self-hosted | |
|---|---|
| Identity | |
| Developer | Technology Innovation Institute |
| Released | May 2025 |
| Status | GA |
| Licence | Open weights (TII Falcon License) |
| Self-hostable | Yes |
| Cost | |
| Input price | Self-hosted |
| Output price | Self-hosted |
| Cached input | Not offered |
| Batch discount | Not offered |
| Free tier | Not offered |
| Capacity | |
| Context window | 256k tokens |
| Capability | |
| Vision in | No |
| Audio in | No |
| Function calling | Yes |
| Structured output | Not offered |
| Extended reasoning | No |
| Web search | Not offered |
| Code execution | Not offered |
| Access | |
| API | No |
| Chat app | Yes |
| Cloud marketplaces | AWS Bedrock Marketplace, Amazon SageMaker JumpStart |
| Fine-tuning | Yes (self-hosted via LoRA/QLoRA - OUMI, Llama-Factory, Axolotl, Unsloth) |