{"slug":"yi-1.5-34b","id":"yi-1.5-34b","type":"model","title":"Yi-1.5 34B","description":"Strong bilingual (English/Chinese) model from 01.AI that competes with much larger models on benchmarks. Excellent reasoning and code generation at a deployable 34B size. Apache 2.0 license.","last_updated":"2026-04-10","last_verified":null,"verification_status":"unverified","markdown_url":"/content/models/yi-1.5-34b.md","html_url":"/models/yi-1.5-34b","api_url":"/api/v1/models/yi-1.5-34b.json","content_hash":"f36c16df029a54e4c586ea6a91a93956de84f9eefcacf6db13e6fed78881c3f8","sha256":"f36c16df029a54e4c586ea6a91a93956de84f9eefcacf6db13e6fed78881c3f8","provider":"01.AI","pricing":{"input":"Free (Apache 2.0)","output":"Free (Apache 2.0)","free":true},"benchmarks":{"reasoning":80,"coding":79,"math":78,"writing":80,"multilingual":82,"speed":85},"tags":["01.ai","open-source","text"],"website":"https://01.ai","release_date":"2025","relationships":{"links":[],"related":[{"id":"cohere-tiny-aya","title":"Cohere Tiny Aya 3.35B","type":"model","html_url":"/models/cohere-tiny-aya","markdown_url":"/content/models/cohere-tiny-aya.md","shared_tags":["open-source","text"],"score":4},{"id":"command-r-plus","title":"Command R+","type":"model","html_url":"/models/command-r-plus","markdown_url":"/content/models/command-r-plus.md","shared_tags":["open-source","text"],"score":4},{"id":"deepseek-r1","title":"DeepSeek R1","type":"model","html_url":"/models/deepseek-r1","markdown_url":"/content/models/deepseek-r1.md","shared_tags":["open-source","text"],"score":4},{"id":"deepseek-v3.2","title":"DeepSeek V3.2","type":"model","html_url":"/models/deepseek-v3.2","markdown_url":"/content/models/deepseek-v3.2.md","shared_tags":["open-source","text"],"score":4},{"id":"falcon-3","title":"Falcon 3","type":"model","html_url":"/models/falcon-3","markdown_url":"/content/models/falcon-3.md","shared_tags":["open-source","text"],"score":4},{"id":"gemma-3","title":"Gemma 3","type":"model","html_url":"/models/gemma-3","markdown_url":"/content/models/gemma-3.md","shared_tags":["open-source","text"],"score":4}],"explicit":{}},"metadata":{"title":"Yi-1.5 34B","type":"model","id":"yi-1.5-34b","provider":"01.AI","model_type":"open-source","release_date":"2025","description":"Strong bilingual (English/Chinese) model from 01.AI that competes with much larger models on benchmarks. Excellent reasoning and code generation at a deployable 34B size. Apache 2.0 license.","last_updated":"2026-04-10","context_window":"32K tokens","website":"https://01.ai","license":"Apache 2.0","modality":["text"],"tags":["01.ai","open-source","text"],"pricing":{"input":"Free (Apache 2.0)","output":"Free (Apache 2.0)","free":true},"benchmarks":{"reasoning":80,"coding":79,"math":78,"writing":80,"multilingual":82,"speed":85},"parameters":"34B (also 6B, 9B variants)","hardware_requirements":"1x RTX 4090 24GB (Q4); 1x A100 40GB (FP16)","best_for":["Bilingual EN/CN applications","Cost-effective self-hosting","Fine-tuning","Research"]},"content_text":"# Yi-1.5 34B\n\nA quietly excellent bilingual model that punches above its weight. Yi-1.5 34B delivers benchmark scores in the 78-82 range across the board on a single RTX 4090, which is remarkable for a dense 34B model. If your workload is English/Chinese and you want Apache 2.0 licensing on consumer hardware, this is the most cost-effective option.\n\nThe scores are evenly distributed -- reasoning 80, coding 79, math 78, writing 80, multilingual 82 -- with no dramatic peaks or valleys. Speed at 85/100 is strong for its size. The 32K context window is the main limitation in a world where 128K is becoming standard, but for most real-world tasks 32K is sufficient.\n\nSelf-hosting is straightforward: an RTX 4090 handles Q4 quantization, or a single A100 40GB runs FP16. The 6B and 9B variants scale down further for lighter deployments. Apache 2.0 means no commercial restrictions. The fine-tuning community is smaller than Llama or Qwen, but the model responds well to LoRA and full fine-tuning for domain-specific tasks.\n\nThe catch is that Yi-1.5 is showing its age. Released in 2025, it predates the current generation of MoE models that deliver more capability per compute dollar. Text-only -- no multimodal support -- further limits its use cases.\n\n**When to pick something else:** Gemma 4's 26B MoE variant offers better benchmarks with multimodal support at similar hardware requirements. For English/Chinese specifically, Qwen 3.5's smaller variants dominate. Yi-1.5 remains relevant mainly for teams already invested in the Yi ecosystem or needing a proven, stable base for fine-tuning.","content_length":2481,"generated_at":"2026-04-24"}