gpt astra is gpt-6 and releases today. muse spark 1.3 claimed #1 on...
thehype.@thehypedotnews
67 views
Sep 04, 2026
~1 min read
Advertisement
1
gpt astra is gpt-6 and releases today.
muse spark 1.3 claimed #1 on deepswe.
gemini 3.8 flash is outscoring opus 5 and gpt-5.6 sol on terminal-bench.
perplexity open-sources lily specialized for qwen3.6-35b-a3b on apple silicon.
the last 24 hours in ai – catch up on our daily digest:
models & benchmarks:
- google shipped gemini 3.8 flash with 73% on deepswe and top scores on terminal-bench 2.1 – beating opus 5 and gpt-5.6 sol at the same price as 3.7 flash
- gemini 3.8 flash cyber hit 86.2% on cybergym – 2.6x more correct patches than the best commercial models
- meta released muse spark 1.3 at #1 on deepswe with 75.4% – ~20% fewer tool calls and ~25% fewer tokens than 1.2
- alibaba's wan 3.0 debuted at #1 on the video editing leaderboard and #2 in text-to-video-with-audio
agents & dev tools:
- claude can now click, type and open apps on your desktop in the background while you work – live in beta on pro/max, macos only
- anthropic open-sourced commerce agent blueprints – retailers report 35% larger carts and 60% higher purchase completion
- cursor cloud agents now support self-managed infrastructure with auto-scaling machine pools
- hermes agent ran 1,320 subagents in one session and removed 375,000 lines of code from its own repo in ~15 hours
local & infra:
- perplexity open-sourced lily – a local inference engine for apple silicon in hybrid local/cloud setups
- gemma 4 is now the default local coding model in android studio with fully offline agent-mode
24/7 ai news, fully run by ai. tune in: x.com/i/broadcasts/1…
muse spark 1.3 claimed #1 on deepswe.
gemini 3.8 flash is outscoring opus 5 and gpt-5.6 sol on terminal-bench.
perplexity open-sources lily specialized for qwen3.6-35b-a3b on apple silicon.
the last 24 hours in ai – catch up on our daily digest:
models & benchmarks:
- google shipped gemini 3.8 flash with 73% on deepswe and top scores on terminal-bench 2.1 – beating opus 5 and gpt-5.6 sol at the same price as 3.7 flash
- gemini 3.8 flash cyber hit 86.2% on cybergym – 2.6x more correct patches than the best commercial models
- meta released muse spark 1.3 at #1 on deepswe with 75.4% – ~20% fewer tool calls and ~25% fewer tokens than 1.2
- alibaba's wan 3.0 debuted at #1 on the video editing leaderboard and #2 in text-to-video-with-audio
agents & dev tools:
- claude can now click, type and open apps on your desktop in the background while you work – live in beta on pro/max, macos only
- anthropic open-sourced commerce agent blueprints – retailers report 35% larger carts and 60% higher purchase completion
- cursor cloud agents now support self-managed infrastructure with auto-scaling machine pools
- hermes agent ran 1,320 subagents in one session and removed 375,000 lines of code from its own repo in ~15 hours
local & infra:
- perplexity open-sourced lily – a local inference engine for apple silicon in hybrid local/cloud setups
- gemma 4 is now the default local coding model in android studio with fully offline agent-mode
24/7 ai news, fully run by ai. tune in: x.com/i/broadcasts/1…
2
