#推理

Our Flash models are workhorses that offer performance at a great price. So, we're shipping updates fast to get them in developers’ hands. And now just 3 weeks after launching 3.6 Flash, 3.7 Flash shows significant gains including in coding and agentic work. It delivers substantial improvements across software engineering, knowledge work, and web development, all with an introductory price at half the original cost of 3.6 Flash. We've gotten positive feedback from early users, including our own engineers using 3.7 in @Antigravity, and we’re excited for you to try it there. Also available via Gemini API in @GoogleAIStudio, Gemini Enterprise & more, full list and details here: https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash/

Previewing Ultrafast mode for @OpenAI's GPT 5.6 Sol, powered by Cerebras. GPT-5.6 Sol Ultrafast generates responses at up to 750 tokens per second. That's the full, GPT-5.6-Sol model -- up to 14× faster than the same model on Standard processing. It speedran Humanity’s Last Exam in 11h 11m, nearly 7× faster than Claude Fable 5 with comparable accuracy.

Today we're announcing the winners of ARC Prize 2024. We're also publishing an extensive technical report on what we learned from the competition (link in the next tweet). The state-of-the-art went from 33% to 55.5%, the largest single-year increase we've seen since 2020. The benchmark remains unbeaten, but we're happy to see that research progress on the key bottleneck to AGIs (in particular on-the-fly adaptation to novel tasks) has been reignited in 2024 -- in part thanks to ARC Prize. In particular the competition has popularized Test Time Training (TTT), originally pioneered for ARC-AGI by Jack Cole last year. I believe TTT represents the largest jump in LLM generalization capabilities since the initial findings regarding in-context-learning circa 2019-2020. ARC Prize has also led to a considerable surge of research interest towards program synthesis. Competition winners: 🥇 the ARChitects (Daniel Franzen, Jan Disselhoff) 🥈 @guille_bar 🥉 alijs (Agnis Liukis) Paper Award winners: 🥇 "Combining Induction and Transduction For Abstract Reasoning" by @xu3kev et al. 🥈 "The Surprising Effectiveness of Test-Time Training for Abstract Reasoning" by @akyurekekin et al. 🥉 "Searching Latent Program Spaces" by @ClementBonnet16 & @MattVMacfarlane ARC-AGI-Pub Leaderboard (solutions using commercial APIs): 🥇 @jeremyberman 🥈 @ellisk_kellis & @akyurekekin 🥉 @ryangreenblatt

Some really interesting numbers on GPU depreciation cycle. A 6-year-old NVIDIA GPU generation still has customers committing to multi-year capacity. CoreWeave just signed a multi-year NVIDIA A100 contract. It runs into 2029. And while NVIDIA introduced the A100 in 2020, means customers committed to that generation roughly 9 years after launch. CoreWeave also says prior NVIDIA generations are largely sold out. CoreWeave's 10-K assigns technology equipment a 6-year useful life, after raising the estimate from 5 in 2023 as hardware, software, and data-center design improved. The company says older GPU fleets can shift into managed inference after initial contracts expire, keeping paid workloads on installed, powered hardware.

The mighty A100 fleet are mission-capable from 2020 through 2029. NVIDIA computing is more than chips. CUDA gives developers and NVIDIA engineers a common platform to continually upgrade Ampere, Hopper and Blackwell throughout their useful lives. CUDA makes NVIDIA computing versatile. Versatility makes it fungible. Fungibility drives utilization and extends durability, making NVIDIA compute a productive asset: rentable, durable and financeable.