Daily AI RoundupsTuesday, 20 October 2026

DeepSeek releases DeepInfer’s offline inference core. Dream Labs opens its Dream-1 robot-control API beta at $3/hour. Inception reports Mercury 3 generates 1,068 tokens/second on NVIDIA GPUs. MiniMax releases M3.1 Flash at $0.15 input/$0.60 output per million tokens. OpenAI reports an 87.4% average pass rate for GPT-6 Astra on its 20-language coding evaluation. Amazon announces TorchNeuron’s October 20 public beta. Cloudflare previews Clairic for extracting structured data. Skyfall AI previews SKY-001 for engineering design. Wiz reports AI-driven cloud attacks. Meta explains its 14-agent workflow. Cloudflare reports a 68% reduction in code-review inference costs. Tenstorrent raises $1.5 billion at a $6.2 billion valuation. DHS publishes its DeepSeek order. Nscale and Anthropic reportedly discuss another $45 billion compute deal.

• AI models
• robotics
• inference
• developer tools
• AI security
• AI policy
• AI infrastructure

Stories in this roundup

1. 01
   DeepSeek open-sources DeepInfer’s offline inference engine for DeepSeek-V4.1-Flash on NVIDIA and Huawei Ascend hardware.DeepSeek
2. 02
   Dream Labs opens Dream-1’s robot-control API beta at $3/hour.Dream Labs
3. 03
   Inception reports its new Mercury 3 diffusion model generates 1,068 tokens/second on NVIDIA GPUs.Inception
4. 04
   MiniMax releases M3.1 Flash at $0.15 input/$0.60 output per million tokens.MiniMax
5. 05
   OpenAI reports GPT-6 Astra’s 87.4% average pass rate on its 20-language coding test, versus GPT-5.6 Sol’s 26.8%.OpenAI
6. 06
   Amazon announces TorchNeuron’s October 20 public beta for native PyTorch on Trainium2 and Trainium3.Amazon / PyTorch
7. 07
   Cloudflare previews Clairic for extracting structured data in a single pass on Workers AI.Cloudflare
8. 08
   Skyfall AI previews SKY-001, claiming 51–125x faster CAD/BIM generation at 150–210x lower cost than tested frontier models.Skyfall AI
9. 09
   Wiz links an AI-driven cloud attacker’s source IP to scans of over 600 customer environments.Wiz
10. 10
    Meta reports 14 specialist agents cut manual data-column preparation from 3–4 days to under one hour.Meta
11. 11
    Cloudflare reports cutting its AI code-review inference bill by 68% through prompt caching and model routing.Cloudflare
12. 12
    Tenstorrent raises $1.5 billion at a $6.2 billion post-money valuation.Tenstorrent
13. 13
    DHS publishes its order excluding DeepSeek-based products from civilian federal procurement and IT systems.DHS / Federal Register
14. 14
    Nscale and Anthropic reportedly discuss another $45 billion cloud-computing deal.Bloomberg via TechCrunch