Chutes

A lot of noise around Chutes right now. Here's Jon's take on our position, plus a first look at what comes next. $TAO #SN64 https://t.co/SFjPufWNCf

We're going to Malaysia next week. Two hackathons, both starting Monday May 18. We're sponsoring both. The first is APU AIC's AI Marathon at Asia Pacific University. Students building AI solutions for real problems, with workshops and industry sessions running alongside it. We're a main sponsor, so participants get full access to the Chutes API and model catalog. → http://ai-marathon-2026.vercel.app The second is Chutes Hack Malaysia with Nyala Labs. Five days, May 18 through 22, at Infinity8 Reserve in Subang Jaya. Free to enter, open to high school students, university students, and working professionals. You can also join online. Workshops on agentic workflows and on-chain inference every day. Prizes include Chutes credits, merch, and a separate award for the best app that uses Sign in with Chutes (our OAuth that lets your users bring their own compute). → Register: http://luma.com/gdre3p9z → Discord: http://discord.gg/MmncAPW9Wb We'll be on the ground all week. If you're a founder or team in SEA and want to grab coffee or talk shop in person, DM us. Happy to make time. Who's coming?

What does your compliance team need to see before client data can hit a cloud AI API? Probably this. Chutes runs Trusted Execution Environments on Intel TDX combined with NVIDIA Confidential Compute mode. Your prompt is decrypted inside a sealed region of the GPU chip, processed there, and re-encrypted before it leaves. The host operator running the box has no read path into that region because the isolation is enforced at the silicon level. Threat model in plain terms: this protects against any privileged user on the GPU operator side, including infrastructure staff and root-level host access. TLS and AES-256 , block the Chutes load balancers from seeing plaintext on the way to the enclave. For workloads that require an even greater level of security, our open-source E2EE proxy adds post-quantum encryption connected directly to a verified TEE instance. Total verifiable privacy from end to end. 16 models on the catalog run inside TEEs today. Each one is flagged with confidential_compute: true in the API response. Forward this to whoever owns your data policy. http://chutes.ai/docs

What if someone found a way to make decentralized large-scale MoE training actually work… while simultaneously building an architecture that drastically reduces VRAM requirement and increases inference throughput? 👀

What can a 27B dense model actually do in 2026? Qwen3.6-27B TEE on Chutes. Apache 2.0. Multimodal. 262K context, extensible to 1M. The benchmarks against Qwen's own previous flagship (397B MoE): SWE-Bench Verified 77.2 vs 70.4 SWE-Bench Pro 53.5 vs 48.4 GPQA Diamond 87.8 vs 84.1 A 27B dense model beating a 397B MoE on every coding eval the vendor publishes. $0.50 in / $2.00 out. Inside a TEE. Run 5M in + 5M out per day on Chutes for just $12.50. What workloads are you still running on bigger models out of habit? https://chutes.ai/app/chute/7aa5e899-c0ba-5482-af48-d3f31d635c9f

A lot of noise around Chutes right now. Here's Jon's take on our position, plus a first look at what comes next. $TAO #SN64 https://t.co/SFjPufWNCf

We're going to Malaysia next week. Two hackathons, both starting Monday May 18. We're sponsoring both. The first is APU AIC's AI Marathon at Asia Pacific University. Students building AI solutions for real problems, with workshops and industry sessions running alongside it. We're a main sponsor, so participants get full access to the Chutes API and model catalog. → http://ai-marathon-2026.vercel.app The second is Chutes Hack Malaysia with Nyala Labs. Five days, May 18 through 22, at Infinity8 Reserve in Subang Jaya. Free to enter, open to high school students, university students, and working professionals. You can also join online. Workshops on agentic workflows and on-chain inference every day. Prizes include Chutes credits, merch, and a separate award for the best app that uses Sign in with Chutes (our OAuth that lets your users bring their own compute). → Register: http://luma.com/gdre3p9z → Discord: http://discord.gg/MmncAPW9Wb We'll be on the ground all week. If you're a founder or team in SEA and want to grab coffee or talk shop in person, DM us. Happy to make time. Who's coming?

What does your compliance team need to see before client data can hit a cloud AI API? Probably this. Chutes runs Trusted Execution Environments on Intel TDX combined with NVIDIA Confidential Compute mode. Your prompt is decrypted inside a sealed region of the GPU chip, processed there, and re-encrypted before it leaves. The host operator running the box has no read path into that region because the isolation is enforced at the silicon level. Threat model in plain terms: this protects against any privileged user on the GPU operator side, including infrastructure staff and root-level host access. TLS and AES-256 , block the Chutes load balancers from seeing plaintext on the way to the enclave. For workloads that require an even greater level of security, our open-source E2EE proxy adds post-quantum encryption connected directly to a verified TEE instance. Total verifiable privacy from end to end. 16 models on the catalog run inside TEEs today. Each one is flagged with confidential_compute: true in the API response. Forward this to whoever owns your data policy. http://chutes.ai/docs

What if someone found a way to make decentralized large-scale MoE training actually work… while simultaneously building an architecture that drastically reduces VRAM requirement and increases inference throughput? 👀

What can a 27B dense model actually do in 2026? Qwen3.6-27B TEE on Chutes. Apache 2.0. Multimodal. 262K context, extensible to 1M. The benchmarks against Qwen's own previous flagship (397B MoE): SWE-Bench Verified 77.2 vs 70.4 SWE-Bench Pro 53.5 vs 48.4 GPQA Diamond 87.8 vs 84.1 A 27B dense model beating a 397B MoE on every coding eval the vendor publishes. $0.50 in / $2.00 out. Inside a TEE. Run 5M in + 5M out per day on Chutes for just $12.50. What workloads are you still running on bigger models out of habit? https://chutes.ai/app/chute/7aa5e899-c0ba-5482-af48-d3f31d635c9f