{"slug": "together-ai", "name": "Together AI", "category": "ai", "category_name": "AI models and assistants", "description": "", "recommenders": 3, "commenters": 12, "warners": 0, "score": 5.002, "rising": 3.0, "first": "2026-01-21", "latest": "2026-10-23", "why": "I\u2019ve mostly settled on using a mixture of open weights models through Together.ai and Fireworks.ai", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 3, "neutral": 5, "compare": 4, "warn": 0}, "spark": [0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 2], "listed": true, "loved": [], "caveat": "", "short_quote": "I\u2019ve mostly settled on using a mixture of open weights models through Together.ai and Fireworks.ai", "kind": "", "icon": true, "domain": "together.ai", "github_repo": null, "official_url": "https://www.together.ai", "pricing": "", "aliases": ["Together AI", "together.ai", "Together.ai"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 803449650, "item_id": 50215603, "url": "https://news.ycombinator.com/item?id=50215603", "by": "joshheitzman", "date": "2026-10-23", "stance": "recommend", "quote": "the inference from together.ai feels like it might be better", "excerpt": "I use that specific model as my primary LLM for coding and diagnostics on projects in Python, C++, CUDA, and for web research. I'm also using it with my own harness (you mentioned \"in my harness\"), so I'll share two things I found about using open weight model providers. The first is that getting the configuration right isn't a per model thing, its a per model per provider per point in time thing. The point in time t", "context": "", "use_case": "open-weight model inference", "switched_from": "", "story": "Kimi K3.1 scores 55 on Artificial Analysis, placed on the Pareto line"}, {"id": 802641616, "item_id": 50165101, "url": "https://news.ycombinator.com/item?id=50165101", "by": "draxen", "date": "2026-10-20", "stance": "recommend", "quote": "Together.ai, Baseten, and others have already achieved higher throughput", "excerpt": "One of the fascinating dynamics around Chinese open-weights models is the competition among American (and some European) inference providers to deliver the best inference performance on them. They've increasingly beaten the Chinese model makers themselves on inference throughput and latency. Together.ai, Baseten, and others have already achieved higher throughput, lower latency, and lower E2E latency for DeepSeek V4.", "context": "", "use_case": "", "switched_from": "", "story": "DeepInfer: High-performance inference framework for DeepSeek models"}, {"id": 790685681, "item_id": 49417855, "url": "https://news.ycombinator.com/item?id=49417855", "by": "dkersten", "date": "2026-08-24", "stance": "recommend", "quote": "I\u2019ve mostly settled on using a mixture of open weights models through Together.ai and Fireworks.ai", "excerpt": "I stopped using Claude because of this BS. If I pay for a service, I want to know what I\u2019m paying for, I want it to be predictable. Anthropic have been anything but. Flip flopping on model availability, model access behind an opaque filter, their past behaviour of model degradation as they prepared their next model\u2026 these are not signs of a reliable service. I\u2019ve mostly settled on using a mixture of open weights mode", "context": "", "use_case": "hosted open-weights models", "switched_from": "Claude", "story": "Anthropic's best AI model struggles to attract users as cheaper tools thrive"}]}], "caveats": [], "who": [], "tags": [], "use_cases": ["open-weight model inference", "running open weight models", "hosted open-weights models", "llm inference provider", "openrouter alternative", "llm inference hosting"], "compared": [{"slug": "fireworks-ai", "name": "Fireworks AI", "co": 5, "switched_from": 0, "switched_to": 0}, {"slug": "claude", "name": "Claude", "co": 1, "switched_from": 1, "switched_to": 0}, {"slug": "openrouter", "name": "OpenRouter", "co": 2, "switched_from": 0, "switched_to": 0}, {"slug": "deepseek-v4-flash", "name": "DeepSeek V4 Flash", "co": 2, "switched_from": 0, "switched_to": 0}, {"slug": "deepseek", "name": "DeepSeek", "co": 2, "switched_from": 0, "switched_to": 0}, {"slug": "groq", "name": "Groq", "co": 2, "switched_from": 0, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=48275546", "title": "Ask HN: Some Sensible Alternative to Openrouter?", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=50213337", "title": "Kimi K3.1 scores 55 on Artificial Analysis, placed on the Pareto line", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=50163277", "title": "DeepInfer: High-performance inference framework for DeepSeek models", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49946321", "title": "Pop!_OS bans AI-generated code from much of its codebase", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49411102", "title": "Anthropic's best AI model struggles to attract users as cheaper tools thrive", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48196656", "title": "Gemini 3.5: frontier intelligence with action", "mentions": 1}], "sentiment_by_month": {"2025-11": {}, "2025-12": {}, "2026-01": {"neutral": 1}, "2026-02": {"compare": 1}, "2026-03": {"compare": 2}, "2026-04": {}, "2026-05": {"compare": 1, "neutral": 3}, "2026-06": {}, "2026-07": {}, "2026-08": {"recommend": 1}, "2026-09": {}, "2026-10": {"neutral": 1, "recommend": 2}}, "quotes": [{"id": 803449650, "item_id": 50215603, "url": "https://news.ycombinator.com/item?id=50215603", "by": "joshheitzman", "date": "2026-10-23", "stance": "recommend", "quote": "the inference from together.ai feels like it might be better", "excerpt": "I use that specific model as my primary LLM for coding and diagnostics on projects in Python, C++, CUDA, and for web research. I'm also using it with my own harness (you mentioned \"in my harness\"), so I'll share two things I found about using open weight model providers. The first is that getting the configuration right isn't a per model thing, its a per model per provider per point in time thing. The point in time t", "context": "", "use_case": "open-weight model inference", "switched_from": "", "story": "Kimi K3.1 scores 55 on Artificial Analysis, placed on the Pareto line"}, {"id": 802641616, "item_id": 50165101, "url": "https://news.ycombinator.com/item?id=50165101", "by": "draxen", "date": "2026-10-20", "stance": "recommend", "quote": "Together.ai, Baseten, and others have already achieved higher throughput", "excerpt": "One of the fascinating dynamics around Chinese open-weights models is the competition among American (and some European) inference providers to deliver the best inference performance on them. They've increasingly beaten the Chinese model makers themselves on inference throughput and latency. Together.ai, Baseten, and others have already achieved higher throughput, lower latency, and lower E2E latency for DeepSeek V4.", "context": "", "use_case": "", "switched_from": "", "story": "DeepInfer: High-performance inference framework for DeepSeek models"}, {"id": 799262450, "item_id": 49953903, "url": "https://news.ycombinator.com/item?id=49953903", "by": "zzzeek", "date": "2026-10-04", "stance": "neutral", "quote": "I use something like together.ai with an open weight model like glm 5.3", "excerpt": "i use claude max and ive yet to get even 30% into my quota for it. so as long as that kind of deal lasts i dont think much about tokens. if i did have to think about tokens I use something like together.ai with an open weight model like glm 5.3 (which ill sometimes use to review a claude change for something intricate). glm 5.x is just very chatty though", "context": "", "use_case": "running open weight models", "switched_from": "", "story": "Pop!_OS bans AI-generated code from much of its codebase"}, {"id": 790685681, "item_id": 49417855, "url": "https://news.ycombinator.com/item?id=49417855", "by": "dkersten", "date": "2026-08-24", "stance": "recommend", "quote": "I\u2019ve mostly settled on using a mixture of open weights models through Together.ai and Fireworks.ai", "excerpt": "I stopped using Claude because of this BS. If I pay for a service, I want to know what I\u2019m paying for, I want it to be predictable. Anthropic have been anything but. Flip flopping on model availability, model access behind an opaque filter, their past behaviour of model degradation as they prepared their next model\u2026 these are not signs of a reliable service. I\u2019ve mostly settled on using a mixture of open weights mode", "context": "", "use_case": "hosted open-weights models", "switched_from": "Claude", "story": "Anthropic's best AI model struggles to attract users as cheaper tools thrive"}, {"id": 772410560, "item_id": 48275660, "url": "https://news.ycombinator.com/item?id=48275660", "by": "cafkafk", "date": "2026-05-25", "stance": "neutral", "quote": "using them as a downstream provider through openrouter at some point (IIRC).", "excerpt": "Not really, beyond using them as a downstream provider through openrouter at some point (IIRC). It worked? \u00af\\_(\u30c4)_/\u00af", "context": "", "use_case": "LLM inference provider", "switched_from": "", "story": "Ask HN: Some Sensible Alternative to Openrouter?"}, {"id": 772409872, "item_id": 48275617, "url": "https://news.ycombinator.com/item?id=48275617", "by": "EbNar", "date": "2026-05-25", "stance": "neutral", "quote": "I've been reading about together.ai.", "excerpt": "Thanks. Not a fan of Google, TBH... By the way, you're right, I'd prefer a credit-based solution. I've been reading about together.ai. Do you happen to knos how is it, by any chance?", "context": "", "use_case": "OpenRouter alternative", "switched_from": "", "story": "Ask HN: Some Sensible Alternative to Openrouter?"}, {"id": 771188864, "item_id": 48199304, "url": "https://news.ycombinator.com/item?id=48199304", "by": "booty", "date": "2026-05-19", "stance": "neutral", "quote": "Open-source model inference providers (who do not have to bear the cost of training) seem able to do it at much lower prices.", "excerpt": "Prevailing wisdom is that serving LLMs at a profit is achievable... it's when you factor in the cost of training them that prices get astronomical real fast. Open-source model inference providers (who do not have to bear the cost of training) seem able to do it at much lower prices. https://www.together.ai/pricing https://fireworks.ai/pricing#serverless-pricing (scroll down to headline models) Of course, it's possibl", "context": "", "use_case": "LLM inference hosting", "switched_from": "", "story": "Gemini 3.5: frontier intelligence with action"}, {"id": 769448593, "item_id": 48090537, "url": "https://news.ycombinator.com/item?id=48090537", "by": "pants2", "date": "2026-05-10", "stance": "compare", "quote": "Together.ai is a good US provider but then it's 15X more expensive", "excerpt": "- Right, R1 affected markets because the market originally believed your theory, but it doesn't any more, which is why V4 didn't move markets at all. - Sure, you can use US infra providers. Together.ai is a good US provider but then it's 15X more expensive than DeepSeek's Chinese-subsidized pricing. It's really not that attractive at that price point. Anthropic and OpenAI are focused on larger models, but Grok 4.3[1]", "context": "", "use_case": "", "switched_from": "", "story": "Gemini API File Search is now multimodal"}, {"id": 761543585, "item_id": 47596474, "url": "https://news.ycombinator.com/item?id=47596474", "by": "nl", "date": "2026-03-31", "stance": "compare", "quote": "the inference prices somewhere like Fireworks or TogetherAI charges is comparable to what Google/AWS/Azure charge for the same model", "excerpt": "> based entirely on the announcements of the inference providers themselves who also get very cagey when asked to show their work I mean sure, it's self reported. But the inference prices somewhere like Fireworks or TogetherAI charges is comparable to what Google/AWS/Azure charge for the same model an we know they aren't losing money - they have public accounts that show it, eg: https://au.finance.yahoo.com/news/wall", "context": "", "use_case": "AI model inference API", "switched_from": "", "story": "OpenAI closes funding round at an $852B valuation"}, {"id": 757686578, "item_id": 47355411, "url": "https://news.ycombinator.com/item?id=47355411", "by": "vshah1016", "date": "2026-03-12", "stance": "compare", "quote": "we get better performance than big players (588 tok/s vs. Together AI's 298", "excerpt": "Hey HN \u2014 I\u2019m Veer and my cofounder is Suryaa. We're building Cumulus Labs (YC W26), and we're releasing our latest product IonRouter (https://ionrouter.io/), an inference API for open-source and fine tuned models. You swap in our base URL, keep your existing OpenAI client code, and get access to any model (open source or finetuned to you) running on our own inference engine. The problem we kept running into: every in", "context": "", "use_case": "", "switched_from": "", "story": "IonRouter (YC W26) \u2013 High throughput, low cost inference"}, {"id": 754196258, "item_id": 47137266, "url": "https://news.ycombinator.com/item?id=47137266", "by": "jadendhm", "date": "2026-02-24", "stance": "compare", "quote": "Together AI is cheap but slower", "excerpt": "I got tired of surprise AI bills at the end of the month. Every time I built something with AI, I had the same problem: Groq is fast but only has a few models, Together AI is cheap but slower, Fireworks is reliable but more expensive. I ended up managing three API keys, three dashboards, and three unpredictable bills. So I built ApeKey. One endpoint that routes your requests automatically across Groq, Together AI, an", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: ApeKey \u2013 One API with predictable pricing"}, {"id": 747454437, "item_id": 46715902, "url": "https://news.ycombinator.com/item?id=46715902", "by": "promptfluid", "date": "2026-01-21", "stance": "neutral", "quote": "falls back to DeepSeek/Together", "excerpt": "I got tired of building the same AI 'plumbing' (vector memory, routing fallbacks, rate limiting) every time I started a project. I wanted a system that was AI-native from the metal up, not just a script with a prompt. The first Cognitive Substrate: It\u2019s an 8-module ecosystem built on Supabase Edge Functions and a year long development sprint. The Modernizer: This is the core. The system analyzes its own archived code", "context": "", "use_case": "Fallback LLM provider", "switched_from": "", "story": null}]}