SemiAnalysis reports that on AgentX (multi-turn, long-context agent workloads), a new AMD MI355x submission beats B300 on total tokens-per-dollar TCO at lower interactivity ranges, shouting out vLLM, AMD, and LMCache engineers—useful signal for agentic serving hardware choices.
Artificial Analysis shipped Intelligence Index v4.2: adds private agentic knowledge-work eval AA-Briefcase and long-document GDP.pdf, doubles held-out weighting, upgrades grading, and drops saturated GPQA Diamond. Claude Fable 5.1 leads; GPT-6 Astra is second with ~4pt gain over GPT-5.6 Sol and strong output-token Pareto efficiency. A major interim benchmark refresh for model selection and cost/quality tradeoffs.
Perplexity says GPT-6 Astra scored 0.682 on WANDR at about $11.98 per task—the highest of any model they tested. Versus Claude Fable 5.1 it is ~13.5% higher score at ~6.1% lower cost; versus Opus 5 it is ~27.0% higher score at only ~3.3% higher cost.
Cognition says GPT-6 Astra is coming to Devin. On FrontierCode 1.1, Astra is within 0.4 points of Fable 5 at ~64% lower cost, and sets a new SOTA on Cognition’s internal testing benchmark with more comprehensive tests, clearer reports, and better video evidence.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.