perfection is achieved not when there is nothing left to add, but when there is nothing left to remove
idaho
Joined April 2009
- Tweets14.1K
- Following1.1K
- Followers1.8K
- Likes30.5K
kgn ๐จ๐ผโ๐ป/๐บ๐ธ retweeted
Weโve spent decades adapting to vector limitations.
Maybe our tools should adapt to us.
Join the waitlist at zoah.com
1) What
GPT-6 Astra makes significant gains in the Artificial Analysis Coding Agent Index, scoring equal to Fable 5 at lower cost. In the Intelligence Index, it uses fewer tokens than GPT-5.6 Sol for similar performance, but this is outweighed by higher prices
Pricing is 2.5x GPT-5.6 Solโs current prices across the board, up from $4/$20 to $10/$50 per million input/output tokens, with the same 90% discount for cache reads and 25% premium for cache writes.
We see distinct stories across our two flagship Indices. In the Artificial Analysis Coding Agent Index, GPT-6 Astra equals Fable 5 at less than half the cost, driven by significant token efficiency gains. In the Artificial Analysis Intelligence Index, GPT-6 Astra is more token efficient than its predecessor for similar performance, but this is offset by the price increase.
Artificial Analysis Coding Agent Index - key takeaways:
โค Rivals top models: In Codex, GPT-6 Astra scores 67 in the Index - approximately equal to Claude Opus 5 and Fable 5 in Claude Code, and Muse Spark 1.3 in Muse Code. Fable 5.1 in Claude Code leads the Index with a score of 70.
โค 70% more token efficient than GPT-5.6 Sol: GPT-6 Astra sees a substantial improvement in token efficiency, using one third of the tokens compared to GPT-5.6 Sol (max) in the Codex harness, and one fifth of the tokens of Claude Opus 5 (xhigh). Various effort levels of the model occupy the Pareto frontier of token efficiency.
โค Leads Coding Agent Index cost efficiency frontier: At max effort, GPT-6 Astra costs about the same as GPT-5.6 Sol (max) while scoring 2 points higher on the Index. Per task, the model is less than half the cost of Claude Fable 5, for the same score.
Artificial Analysis Intelligence Index - key takeaways:
โค Sits beside GPT-5.6 Sol in Intelligence: GPT-6 Astra scores equal to GPT-5.6 Sol in the Index at 61. This is 5 points lower than Claude Fable 5.1 (max with fallback). The model also trails Metaโs newly released Muse Spark 1.3 (max).
โค ~10% fewer output tokens, offset by price increase: GPT-6 Astra defines a new Pareto frontier for Intelligence Index vs Output Tokens per Task - with a ~10% reduction in token use at max effort compared to GPT-5.6 Sol. However, due to the 2.5x increase in price, the model is 75% more expensive per task than its predecessor at max effort.
โค Hallucinates half as much as GPT-5.6 Sol: GPT-6 Astra sees a large jump in AA-Omniscience, our knowledge and hallucination benchmark. This is driven by a significant decrease in hallucination rate from 92% to 51% at max effort. Unlike some models, this improvement does not come at the cost of accuracy - Astra increased accuracy by 4 points at the same time.
โค ~80 point gain in AA-Briefcase Elo: GPT-6 Astra improves ~80 points in AA-Briefcase, our frontier long-horizon knowledge work evaluation. Models are tested on multi-week projects, with many linked tasks and thousands of source files. Astra sees a significant increase in both rubric scores and Analytical Quality Elo in AA-Briefcase compared to its predecessor. In the other direction, we observe a reduction in Presentation Quality Elo, where GPT-5.6 Sol (max) still leads all models.
โค Mixed progress on other evaluations: The model sees a 6 point gain in Humanityโs Last Exam, a long-standing evaluation with emphasis on mathematics, science, and humanities. This is offset by a drop of ~80 Elo points in GDPval-AA v2 - a benchmark we adapted from OpenAIโs dataset measuring economically valuable tasks across 44 occupations. We also observe 2-3 point regressions on other evaluations across a mix of capabilities, including reductions in ฯยณ-Banking (customer support), SciCode (Python problems in a scientific domain), and AA-LCR (long context reasoning over large documents).
Congratulations @OpenAI and @sama on the launch!
Finally!
Sometimes it's easier to show than tell. We're sure this update will help with that. ๐
GitHub CLI now has a repeatable --attach flag that uploads a local image or video. Reference it inline in an issue, pull request, or comment body.
Available now to all users on GitHub across all plans. ๐
github.blog/changelog/2026-0โฆ
kgn ๐จ๐ผโ๐ป/๐บ๐ธ retweeted
A new vector tool is coming to Opacity. Get on the waitlist to try it first: opacity.com
Canโt wait to see local llm benchmarks on this!
M5 Ultra on the @apple store.
1.2TB/s memory bandwidth and up to 512GB unified memory.
512GB can be ordered in late October, but 256GB is available to pre-order.
kgn ๐จ๐ผโ๐ป/๐บ๐ธ retweeted
M6 just dropped
M6 Mac Mini starts at $899 now ๐ฎ
apple.com/newsroom/2026/08/aโฆ
What does the world look like if AI is running at 3000+ tokens per second by the end of the yearโฆ
Artificial Analysis has measured 3,431 tokens/s on Gemma 4 31B via a private demonstration endpoint for NVIDIA Groq 3 LPX - the highest 100k-context output speed we have measured for Gemma 4 31B on either a private or public endpoint
@nvidia announced today that the Groq 3 LPX โinteractive AI inferenceโ rack is in full-scale production and will enter operation later this year. NVIDIA granted us early access to a private deployment of the rack serving Gemma 4 31B for benchmarking purposes
We ran the standard 1k, 10k and 100k input sequence length prompts that we use for benchmarking serverless API providers. At both the 10k and 100k input sequence lengths, the endpoint delivered a median output speed of approx. 3,400 tokens/s, measured over 50 sequential (single-concurrency) requests. Output speed was maintained between the 10k and 100k input sequence lengths tested, indicating robust long-context inference performance
Congratulations to the @nvidia team on the launch!
kgn ๐จ๐ผโ๐ป/๐บ๐ธ retweeted
Watch this video! Weโre rethinking vector :D
The pen tool hasn't changed since 1987. We're rebuilding it at Opacity.
Early access: opacity.com
kgn ๐จ๐ผโ๐ป/๐บ๐ธ retweeted
๐จ GPT-5.6-Sol Ultrafast mode (750 TPS) is here powered by Cerebras!
Here's How kernels work on Cerebras Chips
cerebras built a chip with 900,000 cores on a single silicon wafer.
CSL (Cerebras Software Language) is a Zig-inspired DSL that gives you direct control over the Wafer-Scale Engine. the SDK is publicly documented and the programming model is genuinely the most alien one i've ever looked at. if NKI felt different, CSL takes it much much further.
in CUDA you write one thread's perspective and launch millions. in CSL, there are no threads, no warps, no shared memory, and no kernel launches. you write code for individual Processing Elements: 900,000 independent cores arranged in a 2D mesh on a single silicon wafer. each PE has its own 48 KB of private SRAM, its own program counter, and a 5-port router connecting it to its 4 neighbors. that's actually it. no DRAM. no HBM. no cache hierarchy. 48 KB is your entire world per PE, where code and data must both fit.
the programming model is dataflow, and this is where it gets a bit more alien ๐ฝ
data moves between PEs as 32-bit messages called wavelets, traveling along virtual channels called colors. when a wavelet arrives at a PE on a specific color, it activates a task (a chunk of code bound to that color at compile time). tasks run to completion, then hardware picks the next activated task. tasks cannot call each other. they can only be activated.
so instead of "launch N threads," you think of it like: "place code on PEs, define routes, let data flow."
the memory model is also very different from GPUs.
on an H100 you get 80 GB of HBM shared across all SMs. on the WSE-3, memory is 48 KB per PE, and there are 900,000 of them! this gives you 44 GB total on-chip SRAM at 21 PB/s aggregate bandwidth (vs 3 TB/s on H100). every access is single-cycle. no coalescing needed. no bank conflicts. no cache misses. but also no way to access another PE's memory. all inter-PE communication is explicit wavelet routing through the fabric.
take a distributed GEMV for example. you would write a layout file that physically routes wavelets across the mesh. two PEs sit side by side. the left PE computes a partial result and routes it eastward. the right PE receives those wavelets from the west and accumulates. routing is defined at compile time. both operations are asynchronous and activate a task when they finish. you're physically routing data across silicon at 1 clock cycle per hop.
cerebras is an extremely technically interesting & alien beast. reports 95-210x speedups over H100 on stencil computations. 3,000 tokens/sec inference on gpt-oss-120B. $10B+ deal with OpenAI for 750 MW of inference infrastructure. very exciting times for alternative accelerators!
@gpuemi technical deep dive
kgn ๐จ๐ผโ๐ป/๐บ๐ธ retweeted
"Ultra" mode is not what you think. I did a deep dive on Codex's newest feature to try and help y'all understand it better.
kgn ๐จ๐ผโ๐ป/๐บ๐ธ retweeted
Most AI design tools are built to skip designers.
Describe what you want, a model generates it, you ship whatever it produces.
We think that's the wrong direction.
Our first State of Design is our read on where this actually goes, after three days at Stripe Sessions.
kgn ๐จ๐ผโ๐ป/๐บ๐ธ retweeted
Meet the new Opacity.
A new look for the next evolution in software development. Visually design, build, and deploy stable software at scale, all from @OpacityHQ.