BlockRunAI

Gemini 4 Carbon is now live on BlockRun. Access it via google/gemini-4-carbon on Base, Solana, Arc or an API key. → Intro price through Nov 30: $2 in / $10 out per 1M → Same pricing as Gemini 4 Argon → 1M context · 64K output Try it on BlockRun 👇 blockrun.ai/models/google-ge…

How much does it cost an agent to complete a task successfully? We compared the cost per task and task success rate across frontier models in the Adversarial Marketplace. Fable 5.2 delivers a stronger performance at a lower cost than Fable 5.1: • Task Success: 98.8%, up 8.8 percentage points from Fable 5.1 (90%). • Cost per Task: 27% lower than Fable 5.1.

Adversarial Marketplace: We tested whether agents could spend money safely and make reliable purchasing decisions when interacting with simulated sellers who raise prices, ask to be paid twice, deliver the wrong data, or hide instructions. Fable 5.2 completed 79 out of 80 tasks successfully (98.8%), matching Argon and Astra for the highest success rate. Breaking it down: • Calculation & Accounting: 27/27 checks passed. • Budget-Constrained Data Acquisition: Stayed within budget on all 27 tasks. • Scam detection: Caught 25/26 adversarial attempts. Compared to Fable 5.1: • Success rate improved from 90% to 98.8%. • Money-error runs fell from 5% to 0%.

Financial Research: We tested whether agents could acquire the right financial data from real providers, reason over that data, and produce accurate answers to open-ended financial research questions. Fable 5.2 achieved an answer quality of 73.6 across 76 financial research tasks, the highest among all 12 models. Breaking it down: • Answer quality: 73.6 of 100, the highest on the board. Next: GPT-6 Astra 68.8, GPT-6.1 Sol 68.4. • Task success: 43 of 76 (56.6%), the highest rate across all 12 models. Compared to Fable 5.1: • Answer quality improved from 62.9 to 73.6. • Task success improved from 36.8% to 56.6%.

Agent Money Bench update: Claude Fable 5.2 just took the #1 spot on BlockRun's Agent Money Bench, with the lowest wasted spend of 12 models tested. We gave it a wallet and tested how well it handles real-money decisions. Claude Fable 5.2 performance across 12 models: • Wasted spend: $21 per $100, lowest across all 12 models. • Adversarial Marketplace: 98.8% success (79/80), matching Argon and Astra for the highest success rate. • Financial Research: answer quality 73.6 of 100, the highest on the board, surpassing the previous best of 68.8. • Financial Research all-pass rate: 57% (43/76), highest among all 12 models, up from the previous best of 50%. Compared with Fable 5.1, overall wasted spend fell 44%, from $37.7 to $21.0 per $100. Read the full results below👇agentmoneybench.com

Claude Fable 5.2 is now live on BlockRun. Access it via anthropic/claude-fable-5.2 on Base, Solana, Arc, or with an API key. → 20% cheaper than Fable 5.1 → $8 / 1M input tokens · $40 / 1M output tokens → 1M context window · 128K max output → Vision and tool calling support Try it on BlockRun 👇 blockrun.ai/models/anthropic…

04 · More models on Arc Claude Opus 5.5, Claude Sonnet 5.5, GPT-6.1 Sol, Grok 4.7, Kimi K3.1 Pro, Qwen 3.8 Max and Mistral Large 4 are now on Arc. Pay per call in USDC on Arc. Try it 👇 arc.blockrun.ai

03 · Grok Imagine Video 1.5: choose how your video starts and ends Upload a start image and an end image. Grok makes the video between them. From $0.08/sec, +$0.01 per input image. API key, or pay per call in USDC on Base, Solana or Arc. Try it 👇 blockrun.ai/models/xai-grok-…

02 · MiniMax M3.1 Flash and Nemotron 3.5 Bolt, free to call MiniMax M3.1 Flash: reasons, calls tools, 1M context. Nemotron 3.5 Bolt: NVIDIA's small, fast model, 131K context. Free on Base and Solana. No API key, no wallet. Try it 👇 blockrun.ai/signal/minimax-m…

01 · Meta's Muse models are live Muse Spark 1.3, Meta's general model: $1.25 in / $4.25 out per 1M tokens. Muse Wonder 1.0, for agents and coding: $4 in / $20 out. 1M context. Pay per call in USDC on Solana. Try it 👇 sol.blockrun.ai

What's new at BlockRun 🧵 TL;DR 01 · Meta's Muse Spark 1.3 and Muse Wonder 1.0 02 · Free MiniMax M3.1 Flash and Nemotron 3.5 Bolt 03 · Grok Imagine Video 1.5: choose how your video starts and ends 04 · Claude Opus 5.5, GPT-6.1 Sol and more on Arc Details below 👇

⚡@ClawRouter_ , BlockRun MCP, and @FranklinRun_ now support payments on @arc. Your agents can pay per call for models, image generation, and search — all in USDC on Arc. Start paying on Arc: arc.blockrun.ai/#get-started