Wealth Engine
AI Benchmarks
-
Mistral Large 4 Ships 1T Params, Costs 4x Rivals
Mistral Large 4 ships 1.05T parameters at $1.36 per million input tokens, scoring 38 on the Intelligence Index — but costing 4x open-weights rivals.
-
Reflection Beam Ships 501B Open Weights, 23B Active
Reflection Beam ships 501B open weights with 23B active, claiming GLM-5.2 parity at 3-4x less inference compute. Weights land later in October.
-
Gemini 4 Argon Leads 12 of 18 Benchmarks at $2 a Million
Gemini 4 Argon leads 12 of 18 benchmarks at $2 per million tokens, but only vetted cyber defenders can use it. Pricing doubles on…
-
Claude Sonnet 5.5 Hits 70.6% on Terminal-Bench With Opus-Grade Limits
Claude Sonnet 5.5 scores 70.6% on Terminal-Bench 4.0 at $2/$10 per million tokens, above Opus 5.5’s 66.4% at half the price, with new cyber…
-
MiniMax M3.1 Flash Ships With 5 Reasoning Tiers, No Price
MiniMax M3.1 Flash shipped inside MiniMax Code with 5 reasoning tiers, a 1M-token window, no model card and no published price. Here is the…
-
Xiaomi MiMo V2.6 Pro Tops Open Weights on a $3 Million Budget
Xiaomi MiMo V2.6 Pro scores 46 on Artificial Analysis and ships MIT open weights after a $2.62M RL run. What the $3 million model…
-
Grok 4.7 Ships at $2 per Million Tokens, Trailing Rivals by 7 Points
Grok 4.7 launches at $2/$6 per million tokens, 80% under Claude Fable 5.1, but scores 46 vs 53 on Artificial Analysis. Pricing, benchmarks, who…
-
Sakana Fugu Max Ships at $2 per Million Tokens, 60% Under Rivals
Sakana Fugu Max launched at $2 per million input tokens, 40-60% below Sonnet 5 and GPT-5.6 Terra. Pricing, benchmarks, and who loses the revenue.
-
Cognition SWE-2 Matches Fable 5.1 at 64% Less Cost on a Chinese Base
Cognition SWE-2 hits 50.0% on FrontierCode 1.1 at 64% lower cost than Fable 5.1, built on Moonshot’s 2.8T Kimi K3. No public API, no…
-
OpenAI Jalapeño Chip Beats Blackwell 1.9x Per Watt — Ships 2027
The OpenAI Jalapeño chip beat Nvidia Blackwell by up to 1.9x per watt in SemiAnalysis tests. Volume ships 2027. What it means for Nvidia…
-
Best Computer Use Agent 2026: Qwen-UI-Agent vs Opus 4.8
Qwen-UI-Agent tops mobile benchmarks but ships no weights. We price every computer use agent you can deploy in 2026 and pick a clear winner.