This is the first model drop on a Saturday? Interesting choice 😄
This is the first model drop on a Saturday? Interesting choice 😄
AISummary
A new model reportedly scores 77.0 on DeepSWE, above the prior 74.2 and Opus 5's 74.0, at about $0.20 per task versus about $12 for Opus 5.
It also beats Opus 5 on AutomationBench, Agents' Last Exam and Apex, but trails it on HLE, Terminal-Bench, ProgramBench, NL2Repo and the visual benchmarks.
The quoted thread describes a 552B-parameter mixture-of-experts model with 196B Engram memory and a 1M context window, priced at $0.30 per million input tokens and $1.20 output at peak.
AIWhy it matters
The quoted thread gives concrete benchmark scores, per-task cost, and memory figures, which show how the model trades off accuracy, cost, and context against Opus 5.
Post on XView on X
Sophia YangVerified on X
@sophiamyang
TL;DR - Strong on agentic tasks, better and cheaper per task: higher scores with shorter outputs. On DeepSWE it gets 77.0 versus 74.2 before, and beats Opus 5's 74.0, at roughly $0.20 per task against ~$12 for Opus 5. - Beats Opus 5 on four benchmarks: DeepSWE, AutomationBench, Agents' Last Exam and Apex. It trails Opus 5 on HLE, Terminal-Bench, ProgramBench, NL2Repo and the visual benchmarks. - The model: a mixture of experts with 552B parameters plus 196B of Engram memory (~763B total), only 8B active during prefill and 16B during decode. 1M context. - Tiny memory per token: KV cache stored at 4-bit, about 890 bytes per token, roughly a quarter of V4-Flash. - Price: $0.30 per million tokens in and $1.20 out at peak, half off-peak.View quoted post on X
Source: Sophia Yang · x.comPublished