MiniMax M3.1 Flash reshapes BuildingBench’s cost–quality frontier! Another open model making moves in 3D 🏗️🔥 🎈 73.3 overall, ranking #12. 🎈 Above SWE-2’s 67.3, taking its place on the Pareto frontier. 🎈 Close to the previous DeepSeek V4.1 Flash’s 72.9—and currently free through OpenRouter! Its best reconstruction hits 88.8 on Melbourne’s Australian Centre for Contemporary Art. How does it look alongside DeepSeek V4.1 Flash 1017 and #1 Opus 5.5? Comparison video below 👇

See DeepSeek’s 3D leap with your own eyes 👀🏗️ DeepSeek V4.1 Flash 1017 🆚 DS V4.1 Flash 🆚 GPT-6 Astra: three buildings, side by side. The new open model improves on 11 of 12 buildings in BuildingBench, jumping from 72.9 → 81.9—closing in on Astra’s 84.3 at less than half the median cost ($3.01 vs $7.05/task). Open models are getting seriously good at building worlds 🔥

Wow🙀🙀 DeepSeek V4.1 Flash 1017 joins BuildingBench’s top five—an exciting leap for open models!! 🔥 We tested its 3D building reconstruction from real photos. An open model delivering this level of spatial reasoning at this cost is exciting to see! 👀 🎈 81.9 overall, up from 72.9 for the previous version. A 9-point jump! 🎈 Slightly higher than Fable 5.1’s 81.4, and essentially level with Sonnet 5.5’s 82.1. 🎈 Estimated median cost: $3.01/task. About 1/11 of Fable 5.1’s cost and 1/15 of Sonnet 5.5’s.

🎉Excited to see two new US open models from this week: 📘Meta’s Muse Spark Open 📗NVIDIA’s Nemotron 3.5 Super VL How well can they build in 3D? 🏗️👀 We tested both on BuildingBench: reconstructing real buildings from photographs. 📘Muse Spark Open (max): 56.9 at an estimated $8.59/task—comparable to Kimi K3’s 54.0. 📗 Nemotron 3.5 Super VL (max): 37.8 at an estimated $0.76/task—among the cheapest tested, with performance near GLM-5.3 Flash’s 33.0. DeepSeek V4.1 Flash remains well ahead of both: 72.9 at an estimated $2.33/task. GLM-5.3 FlashX also scores higher, at 68.0. Neither newcomer reaches the cost–quality frontier. More open models to choose from, and plenty of room to improve spatial reasoning! Both ran in Claude Code at max effort. Costs are estimated median per task.

The completed WorldBox run reaches 230 rooms and 61 gems at 35,000 actions using 973M tokens. It also reaches Astra’s 23-gem result in just 10,929 actions—roughly half as many.

Not just mazes—WorldBox helps agents in Minecraft, too ⛏️ With WorldBox, Opus 5.5 reached more than twice the average progression tier of the base agent in our tests. Astra improved substantially, too. Spatial memory management and sandbox help agents remember where resources are—and put that knowledge to work. From chasing gems to mining diamonds 💎

Watch how GPT‑6 Astra + WorldBox tackle MazeBench, from its first action to gem💎 #61—32,980 live actions and around 1B tokens. Building bridges with boxes, launching across gaps, solving puzzles that span multiple rooms… all while the map fills in. Remember, test, revise, repeat, and grab that gem. WorldBox offers the supporting layer for Astra.

🚨New level unlocked! 🚨 We doubled GPT‑6 Astra’s score on MazeBench with WorldBox, our agentic sandbox for spatial reasoning. Astra: 23 gems 💎💎 Astra+WorldBox: 61 gems 💎💎💎💎💎 MazeBench, by @patience_cave, is a tough test of long-horizon spatial reasoning: 256 interconnected rooms where even Opus 5.5 used 2.2B tokens and collected just 13 of 100 gems. WorldBox provides agents with an agentic sandbox, for managing long-term memory, inferring complex mechanics, and simulating reasoning. Watch the sandbox take shape and the agent tackle increasingly tricky puzzles. 🧵

Gemini 4 Argon makes a big leap in 3D reconstruction 🏗️🚀 On BuildingBench, Argon (high) scores 81.2—up from Gemini 3.8 Flash’s 54.8, placing #6 among all 30 models. Near Fable 5.1 quality at ~1/5 the cost. Near Sonnet 5.5 quality at ~1/7. 👀 💸 Estimated median cost: $6.52 per building 🎨 Appearance score: 94.1, the third-highest 🏛️ Best reconstruction: Royal Exhibition Building, scoring 92.1 It doesn’t move the Pareto frontier: GPT‑6.1 Sol still scores higher at lower cost. But this is a substantial step forward for Gemini. Congrats @GeminiApp team!

Muse Wonder takes on #BuildingBench 🏗️👀 How well can it turn photographs into 3D buildings? Muse Wonder (max) scores 70.3, ranking #12 of 31—roughly level with GPT‑6 Sol. Compared with Muse Spark 1.3: 📈 Score: 68.2 → 70.3 💸 Median cost per building: $5.89 → ~$7.72 📐 Higher geometry and surface scores, but lower appearance scores. A modest gain within statistical uncertainty. The Pareto frontier stays unchanged. Watch its best reconstruction: Helsinki’s Pikkuparlamentti, scoring 81.6 👇

Two models this week: Haiku 5.5 and Mistral Large 4 take on BuildingBench 🏗️👀 🚀 Haiku 5.5 (max) jumps from its predecessor’s 26.7 to 67.1! It beats Sonnet 5 by 13 points at roughly 1/11 the cost: $1.34 vs. $15.03 per building. 🇫🇷 Mistral Large 4 (max) makes its BuildingBench debut at 52.0, using the Claude Code harness. At $33.21 per building, cost-efficiency has room to improve. Haiku 5.5 narrowly misses the Pareto frontier: SWE‑2 scores almost identically and is still free during its promotion. 👀

We gave Opus 5.5 and GPT‑6 Astra a memory of their Minecraft world. Their progress leaped: 🚀 Opus 5.5: +133% 🔥 🚀 Astra: +50% 🔥 Introducing WorldBox, Enactra’s sandbox that gives agents spatial memory: where they’ve explored, where resources are, and what they can return to. Watch Opus turn that memory into a diamond pickaxe 💎 Side-by-side race with and without WorldBox in the thread 👇

GPT‑6.1 Sol and Sonnet 5.5 leap up BuildingBench’s leaderboard! 🏗️ 🥈 Sol reshapes the cost–quality frontier greatly! It scores 85.1, up from GPT‑6 Sol’s 69.6—taking second place while cutting median cost from ~$1.89 to ~$1.70 per building. That edges past GPT‑6 Astra (ultra, 84.3) at 76% lower cost. Just 1.7 points behind Opus 5.5 (max, 86.8), at roughly 1/27 the cost. 🚀 Sonnet 5.5 (max) climbs to #5 with 76.9, up from Sonnet 5’s 54.4—a 22.5-point leap! The gain comes with higher cost: ~$39.99 vs. $15.03 per building. Two big quality jumps. Very different price tags.

SWE‑2 reshapes BuildingBench’s cost–quality frontier 👀🏗️ @cognition's SWE‑2 (max) scores 67.3—just 2.5 points behind GPT‑6 Sol (ultra, 69.8), and ahead of GPT‑5.6 Luna (61.6). Free to use for now, it reshapes the budget end of the cost–quality curve: building in 3D just got more accessible. It's really a good model. Congrats @cognition @ScottWu46 @jaredpalmer ! How much can you build on a $0 model budget? 🚀