OrcaRouter 🐳

Fable 5 is dead. We just resurrected it — cheaper, open and you hold the keys. OpenRouter dropped Fusion 48h ago and broke the internet. We tested it hard. The synthesizer is insane for deep research… but absolute dogshit for coding. So we fixed it. Meet OrcaRouter.ai DSL — the version you actually own. One prompt → fans out to any panel you want → judge + synthesizer → one god-tier answer. But unlike black-box slugs, you control the entire graph in YAML. Fable 5 level intelligence… without waiting for Anthropic to turn it back on 🧵👇

@jietang Amazing, congrats!🎉

We're introducing GLM-5.2, our latest flagship model for long-horizon tasks. It marks a substantial leap in long-horizon task capability over its predecessor GLM-5.1 and, for the first time, delivers that capability on a solid 1M-token context. GLM-5.2's new capabilities include: Solid 1M Context: A solid 1M-token context that stably sustains long-horizon work Advanced Coding with Flexible Effort: Stronger coding capabilities with multiple thinking effort levels to balance performance and latency Improved Architecture: We propose IndexShare, which reuses the same indexer across every four sparse attention layers, reducing per-token FLOPs by 2.9× at a 1M context length. We also improve GLM-5.2’s MTP layer for speculative decoding, increasing the acceptance length by up to 20% Pure Open: An MIT open-source license — no regional limits, technical access without borders Supporting long-horizon tasks starts with making long context engineering-usable: the model must maintain quality across long, messy coding-agent trajectories, not just accept more tokens. A 1M context is easy to claim, but much harder to keep reliable under real engineering pressure. To this end, we substantially expanded 1M-context training for coding-agent scenarios, covering large-scale implementation, automated research, performance optimization, and complex debugging. The result is a long-context system that is not only wide in scope, but solid in execution: a practical substrate for sustained engineering work. This capability is reflected in GLM-5.2's performance on three long-horizon coding benchmarks. FrontierSWE measures whether an agent can complete open-ended technical projects at the scale of hours to tens of hours, spanning systems optimization, large-scale code construction, and applied ML research. On this benchmark, GLM-5.2 trails Opus 4.8 by only 1%, while edging out GPT-5.5 by 1% and Opus 4.7 by 11%. On PostTrainBench, where each agent is given an H100 GPU and evaluated by how much it can improve small models through post-training, GLM-5.2 outperforms both Opus 4.7 and GPT-5.5, ranking second only to Opus 4.8. On SWE-Marathon, an ultra-long-horizon software engineering benchmark covering tasks such as building compilers, optimizing kernels, and developing production-grade services, GLM-5.2 still has room to grow, trailing Opus 4.8 by 13% while remaining second only to the Opus series. Across all three benchmarks, GLM-5.2 is the highest-ranked open-source model, showing that its 1M context has translated into practical long-horizon delivery capability.

@Zai_org Two levels of reasoning effort: GLM-5.2 (max) pushes the limits, while GLM-5.2 (high) strikes a strong balance between performance and token efficiency Weights: huggingface.co/zai-org/GLM-5.2

Le Chaton Fat is hilarious… But @Zai_org just shipped GLM-5.2 with real 1M usable context, MIT incoming, and agent coding superpowers. Memes are fun. Shipping is better. API live on OrcaRouter: orcarouter.ai/models/z-ai/gl…

🔥OrcaRouter is now available in SGLang @sgl_project @radixark @lmsysorg Route across 200+ AI models with a single endpoint with programmable routing, unified billing, Agent Firewall, Guardrails and more.

@probiex007 That’s cool, which models did you use for that?

🤯 This is a website, a simple web-based game built with WebGL and Three.js. Website: messenger.abeto.co It's honestly surprising how far web development has come.

@gonlenidefi @WesRoth Choosing the right arbiter is key here, synthesize doesn’t work in coding

@WesRoth routing DSL for multi-model AI feels like a new middleware category curious how the synthesizer handles conflicting outputs in real time

OrcaRouter released a routing DSL that lets developers build customizable multi-model AI systems using simple YAML configurations. A single prompt can be sent to several models in parallel. A judge or synthesizer then compares the responses and returns one final answer. Developers can control the full workflow, including: • Which models participate • Which providers handle requests • Which model acts as the judge • When expensive multi-model routing is activated • Whether code must pass tests before being returned • Shadow testing, canary deployments, and auditing

@StillWorthFig We do support that already

输入提示词的难度分级，才是决定了它是去往“大而全”还是“小而美”。

@Dudechillkar @opencode Are you a Go plan user? Which models do you need the most or its not relevant?

I think you guys should do what @opencode did and launch a Go plan, i checked out what you guys are doing and i like it. Keep going 💪

🔥OrcaRouter is now on @apify! One endpoint for OpenAI, Claude, Gemini, Grok, DeepSeek, Qwen and more + programmable routing. Zero extra code. Saves tokens → apify.com/continuum-ai-c…

@tridentasia You can achieve on par performance to fable on coding, we will also share some evals we’ve done

@OrcaRouter There is no evidence that fusion model council beats fable on coding benchmarks.

Fable 5 is dead. We just resurrected it — cheaper, open and you hold the keys. OpenRouter dropped Fusion 48h ago and broke the internet. We tested it hard. The synthesizer is insane for deep research… but absolute dogshit for coding. So we fixed it. Meet OrcaRouter.ai DSL — the version you actually own. One prompt → fans out to any panel you want → judge + synthesizer → one god-tier answer. But unlike black-box slugs, you control the entire graph in YAML. Fable 5 level intelligence… without waiting for Anthropic to turn it back on 🧵👇

@deified_duck They are, that’s why we support 5 arbiters and you can choose your own model as judge

@OrcaRouter Aren't models biased towards their own work ?

@OrcaRouter Ending black-box dependency and putting full graph control in devs' hands via YAML is pure genius. Game-changing move! 🚀

@OrcaRouter saved me from another night of screaming at broken python scripts 🖥️

@youpmelone Check out: docs.orcarouter.ai/integrations/o…

@OrcaRouter if it doesnt work with CLI then why

@Tanmoy_ai Try it out

@OrcaRouter I want to try it

@PixelRainbowNFT @junkim_100 We can also natively integrate with Hermes agent, they're however not reviewing our PR for weeks: github.com/NousResearch/h… @NousResearch

@junkim_100 @OrcaRouter Built this for Hermes agent over the weekend. Need to test it more before I open it up.

@ProxyAyush Why new repos if you can build fast directly inside of OrcaRouter?

@OrcaRouter This is another open-source I made for Antigravity, ClaudeCode and Codex x.com/i/status/20665… #fable #anthropic #claudecode #OpenAI #AI #mythos

@giannoklein No deployment anymore with routing DSL, just change base url and go

@OrcaRouter Beat you to it but the bigger name always wins in the end: github.com/GiannoKlein9/H…

@OrcaRouter totally amazing