正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Introducing d3-turbo: our newest state-of-the-art multimodal foundation decision model⚡ #1 on both Jev Decision Index boards. It points(box), follows and sees images in your questions. StateOnce brings 15x faster than d3 flash 🤗 #vLLM #Jev #Decision

Xunzhuo

1,195 subscribers

21,223 次观看 • 26 天前 •via X (Twitter)

科学技术 #vLLM #Jev #Decision

Anya Rossi• Live Now

Private livecam show

26 条评论

Xunzhuo 的头像

Xunzhuo26 天前

🆓 Try it free: One click, no account. The free API now serves d3-turbo, and vllm-sr/d3 names the same model: code that calls d3 is already on turbo. Open weights, Apache-2.0 🤗

Xunzhuo 的头像

Xunzhuo26 天前

📦 New: Box. Ask where something is, get one box with a calibrated confidence. Four questions about one photo, one request: the frisbee 0.99, the dog's tail 0.98, the cone 1.00. 0.90 accuracy at IoU 0.5 · confidence AUROC 0.96

Xunzhuo 的头像

Xunzhuo26 天前

🧪 Real network, real numbers: one curl to the free API, 64 questions about one photo, 64 answers back in 0.75 s. Every answer comes with its probability, and not a single token of text is generated.

Xunzhuo 的头像

Xunzhuo26 天前

Checkout the d3 turbo HF space 🤗 collected some cool demos 👻

Diego Alejandro 的头像

Diego Alejandro26 天前

Down to the detail 🤌

Xunzhuo 的头像

Xunzhuo26 天前

Nice!!!

Xunzhuo 的头像

Xunzhuo26 天前

🖼️ New: images inside questions and options. "Which picture shows the same kind of animal?" Put the pictures right in the options; only that question sees them. 0.96 accuracy, +0.31 over d3.

Xunzhuo 的头像

Xunzhuo26 天前

⚡ State once, answer many. d3-turbo reads your state once and answers every question from it. A testcase like 255 questions about one photo: 2.12 s, vs 32.94 s for d3-flash. 15.5× faster.

Xunzhuo 的头像

Xunzhuo26 天前

🎯 New: Track. Follow one object through a whole video, 1 or 2 updates per second, each with the probability it's visible. Artemis I, from the launch pad into the sky: 28 boxes, one request. Presence F1 0.97 · success AUC 0.80

Diego Alejandro 的头像

Diego Alejandro26 天前

Beautiful video, beautiful model

Xunzhuo 的头像

Xunzhuo26 天前

Cheers!🥂

Sophia AI & Tool Expert 的头像

Sophia AI & Tool Expert26 天前

Huge congrats on d3 turbo #1 on both Jev boards with 15x faster speed is an incredible leap

Xunzhuo 的头像

Xunzhuo26 天前

🌍 Every scene, a decision. Bruegel and Seurat, Earthrise, Jupiter, Curiosity on Mars, a ptarmigan hidden in the snow, a rocket leaving the pad: 37 scenes, and every box, track and pick is a real d3-turbo answer, drawn as returned.

Aina Ai | Tools & Updates 的头像

Aina Ai | Tools & Updates26 天前

State of the art and 15x faster #1 spot well deserved

Alex Ai | Tools & Updates 的头像

Alex Ai | Tools & Updates26 天前

d3-turbo topping decision index with 15x speed gain is impressive

tax check pls 的头像

tax check pls26 天前

'#1 on Jev Decision Index' plus 'StateOnce 15x faster than d3 flash' is a lot of numbers with no eval card. Which benchmark, what sample size? Every model release claims a #1 board these days.

Zara Tech & Tools 的头像

Zara Tech & Tools26 天前

Fastest decisions sharpest vision

ZenithAi 的头像

ZenithAi26 天前

Number one ranking with massive acceleration plus visual grounding is phenomenal

Alice The Ai Expert 的头像

Alice The Ai Expert26 天前

Awesome launch #1 on both boards and 15x faster is incredible, congrats on d3-turbo!

NOOR TECH 的头像

NOOR TECH26 天前

d3-turbo: new multimodal decision model, #1 on Jev Decision Index, with pointing and 15x speed over d3 flash

Evia AI 的头像

Evia AI26 天前

Massive achievement leading benchmarks with incredible efficiency

Muscle AI 的头像

Muscle AI26 天前

Congrats on the launch. Hard question: what does a decision model do that a VLM with a tool harness can't? The pitch is cheaper, faster agent loops, but I'd want head-to-head task success vs a general VLM before 'decision model' earns its own category.

Nova Tech 的头像

Nova Tech25 天前

State of art breakthrough incredible speed and vision

Diakou 的头像

Diakou26 天前

How is this possible

飞叔的头像

飞叔26 天前

For the box-pointing capability, how does it handle several visually similar targets when the instruction is ambiguous? Returning uncertainty or asking which one would be useful alongside the speed improvement.

Xunzhuo 的头像

Xunzhuo26 天前

📈 Better everywhere: +6.2 on the public suite over d3, ahead in all five areas. Arts +9.6 · Retrieval +9.1 · Language +6.4 · Tools +4.3 · Knowledge +3.7 Plus long context: up to 131,072 tokens per question.

相关视频

Introducing Decision 3.0: towards open multimodal foundation decision models ⚡ Open and state-of-the-art from 0.6B to 27B: #1 on both Jev Decision Index boards in our evaluation, on a new Pareto frontier. HF collection 🤗: #OpenSource #Jev #LLM

Introducing Decision 3.0: towards open multimodal foundation decision models ⚡ Open and state-of-the-art from 0.6B to 27B: #1 on both Jev Decision Index boards in our evaluation, on a new Pareto frontier. HF collection 🤗: #OpenSource #Jev #LLM

Xunzhuo

141,794 次观看 • 1 个月前

Meet d3-ultra: our newest open 27B multimodal decision model. #2 of 230 on the Jev Decision Index text board, +3.03 over d3. Text, images, video: it boxes what it finds and tracks it through time. Open weights (Apache-2.0): #Jev #Decision

Meet d3-ultra: our newest open 27B multimodal decision model. #2 of 230 on the Jev Decision Index text board, +3.03 over d3. Text, images, video: it boxes what it finds and tracks it through time. Open weights (Apache-2.0): #Jev #Decision

Xunzhuo

12,534 次观看 • 19 天前

Introducing Decision 2.0: our newest state-of-the-art decision models, open in every size from 0.6B to 27B ⚡ HF collection 🤗: #OpenSource #Decision #Jev #LLM

Introducing Decision 2.0: our newest state-of-the-art decision models, open in every size from 0.6B to 27B ⚡ HF collection 🤗: #OpenSource #Decision #Jev #LLM

Xunzhuo

147,651 次观看 • 1 个月前

Introducing Drex - a lightning-fast decision model built to rival Jev. #1 on the Decision Index. (official scores tbd) winning 23 out of 40 benchmarks Architecture: Small Diffusion Model with RLAF Price: $0.04 per 1M input tokens (cheaper than Jev) Latency: less than a second. Sign up now for 250M welcome credits. Open weights and the full tech report are coming very soon. 👀 🦖 #jev #nace #drex

Introducing Drex - a lightning-fast decision model built to rival Jev. #1 on the Decision Index. (official scores tbd) winning 23 out of 40 benchmarks Architecture: Small Diffusion Model with RLAF Price: $0.04 per 1M input tokens (cheaper than Jev) Latency: less than a second. Sign up now for 250M welcome credits. Open weights and the full tech report are coming very soon. 👀 🦖 #jev #nace #drex

Nace AI

639,893 次观看 • 1 个月前

Introducing the Decision Index 0.1 ⚖️ a rigorous leaderboard comparing jev with 30+ open weights decision models 35+ benchmarks. asking 130K questions to each model testing knowledge 🧠, automation ⚙️, understanding 🤔and even creativity 🎨

Introducing the Decision Index 0.1 ⚖️ a rigorous leaderboard comparing jev with 30+ open weights decision models 35+ benchmarks. asking 130K questions to each model testing knowledge 🧠, automation ⚙️, understanding 🤔and even creativity 🎨

apolinario (poli)

44,399 次观看 • 1 个月前

Jev Engineering changes the unit of optimization from “agent” to “decision.” that sounds small, but it changes the architecture. and up to 193x faster and 444x cheaper in tests. normal multi-agent stack: agent → prompt → tool → output Jev-style stack: shared state → decision boundary → selected branch → execution → measurement → state update now you can optimize the actual control points: latency per decision cost per fork confidence threshold retry policy branch survival termination condition the agents underneath can change. the models can change. the tools can change. the decision layer stays measurable. that’s why Jev Engineering feels more like systems engineering than prompt engineering.

Jev Engineering changes the unit of optimization from “agent” to “decision.” that sounds small, but it changes the architecture. and up to 193x faster and 444x cheaper in tests. normal multi-agent stack: agent → prompt → tool → output Jev-style stack: shared state → decision boundary → selected branch → execution → measurement → state update now you can optimize the actual control points: latency per decision cost per fork confidence threshold retry policy branch survival termination condition the agents underneath can change. the models can change. the tools can change. the decision layer stays measurable. that’s why Jev Engineering feels more like systems engineering than prompt engineering.

Ricker

22,469 次观看 • 1 个月前

Jev is cool, but like any foundation model it needs to be calibrated to your decision criteria. We launched jev-align: an open-source CLI to quickly teach Jev what good and bad looks like using GEPA. Try it out!

Jev is cool, but like any foundation model it needs to be calibrated to your decision criteria. We launched jev-align: an open-source CLI to quickly teach Jev what good and bad looks like using GEPA. Try it out!

Seth Kimmel

129,932 次观看 • 1 个月前

Today, we are announcing Nimble-v4 and Nimble-v4-max which are pushing the frontier of decision models, outperforming Jev and other variants. On the Decision Index public scores, nimble-v4-max and nimble-v4 score +10 and +6 against Jev 1.13, making nimble-v4-max the #1 model. Beyond benchmarks, we have done a lot of internal evals, and see meaningful improvements: 43% fewer mistakes than Jev, 50pp improvement on model routing decisions, being great at coding use cases, fact checking etc. The models are available via our API service. Go build nimbly:

Today, we are announcing Nimble-v4 and Nimble-v4-max which are pushing the frontier of decision models, outperforming Jev and other variants. On the Decision Index public scores, nimble-v4-max and nimble-v4 score +10 and +6 against Jev 1.13, making nimble-v4-max the #1 model. Beyond benchmarks, we have done a lot of internal evals, and see meaningful improvements: 43% fewer mistakes than Jev, 50pp improvement on model routing decisions, being great at coding use cases, fact checking etc. The models are available via our API service. Go build nimbly:

Mahesh Sathiamoorthy

83,826 次观看 • 1 个月前

Jev Engineering is what turns an agent stack into an actual control system and moves the expensive model out of every decision loop. and up to 193x faster and 444x cheaper in tests. the model shouldn’t decide everything. in this setup: request → structured state → Jev router → cheapest capable model → worker → relevance / approval checks → execution gate → tool the important part is that the decision layers don’t generate prose. they route → score → block → approve. so expensive models only get called when the task actually needs them. that’s the point of Jev Engineering: separate reasoning from decision-making, then make the decision layer measurable, cheap and fast. full breakdown in the article below ↓

Jev Engineering is what turns an agent stack into an actual control system and moves the expensive model out of every decision loop. and up to 193x faster and 444x cheaper in tests. the model shouldn’t decide everything. in this setup: request → structured state → Jev router → cheapest capable model → worker → relevance / approval checks → execution gate → tool the important part is that the decision layers don’t generate prose. they route → score → block → approve. so expensive models only get called when the task actually needs them. that’s the point of Jev Engineering: separate reasoning from decision-making, then make the decision layer measurable, cheap and fast. full breakdown in the article below ↓

Ricker

183,199 次观看 • 1 个月前

Jev Engineering is the layer most agent stacks are still missing. state → decision → action → verification → next state up to 193x faster and 444x cheaper in our tests. everyone is optimizing the model. Jev Engineering optimizes what happens between model calls. every agent eventually hits the same fork: state → multiple possible actions → one route has to win instead of sending every decision back to an expensive LLM call, Jev turns that into a dedicated layer: so the loop becomes: LLM → reason Jev → decide tools / agents → execute state → update Jev → decide again once decisions become their own engineering layer, you can benchmark, batch and verify them separately. full blueprint in article ↓

Jev Engineering is the layer most agent stacks are still missing. state → decision → action → verification → next state up to 193x faster and 444x cheaper in our tests. everyone is optimizing the model. Jev Engineering optimizes what happens between model calls. every agent eventually hits the same fork: state → multiple possible actions → one route has to win instead of sending every decision back to an expensive LLM call, Jev turns that into a dedicated layer: so the loop becomes: LLM → reason Jev → decide tools / agents → execute state → update Jev → decide again once decisions become their own engineering layer, you can benchmark, batch and verify them separately. full blueprint in article ↓

Ricker

116,701 次观看 • 1 个月前

What is a decision model? Jev by TypeSafe AI answers yes/no and multiple-choice questions, with a confidence score. Much of software development are a sequence of decisions, and Jev is 10x cheaper and faster than an LLM. Let’s understand this through practical examples:

What is a decision model? Jev by TypeSafe AI answers yes/no and multiple-choice questions, with a confidence score. Much of software development are a sequence of decisions, and Jev is 10x cheaper and faster than an LLM. Let’s understand this through practical examples:

OpenRouter

169,880 次观看 • 1 个月前

Jev x GPT-6 Astra x H3 Max I created a complex decision making game with Astra. It is about keeping the city on a whale alive with decisions. - GPT 6 Astra designed the world - Jev chose the actions - H3 Max Turbo on fal turned one decision from each round into video All of the Jev decision making and video generations took only 5 minutes for 264 clips

Jev x GPT-6 Astra x H3 Max I created a complex decision making game with Astra. It is about keeping the city on a whale alive with decisions. - GPT 6 Astra designed the world - Jev chose the actions - H3 Max Turbo on fal turned one decision from each round into video All of the Jev decision making and video generations took only 5 minutes for 264 clips

gokaygokay

21,821 次观看 • 1 个月前

Introducing THIS / THAT Model 1.0. Trained by FLock. Not another chatbot. A best-in-class open-source model built to decide. 94.1% success rate on our 68-question decision test, outperforming GPT-5.6, Deepseek-4.1-flash, Kimi K3, Jev, and more. Just 1.9B parameters. Runs locally. Measured decision latency of ~31 ms on an RTX 5080 laptop GPU. We’re putting decentralized AI to work: giving developers and AI agents decision-making capabilities they can run on their own hardware. Your model. Your machine. Not every decision needs the cloud.

Introducing THIS / THAT Model 1.0. Trained by FLock. Not another chatbot. A best-in-class open-source model built to decide. 94.1% success rate on our 68-question decision test, outperforming GPT-5.6, Deepseek-4.1-flash, Kimi K3, Jev, and more. Just 1.9B parameters. Runs locally. Measured decision latency of ~31 ms on an RTX 5080 laptop GPU. We’re putting decentralized AI to work: giving developers and AI agents decision-making capabilities they can run on their own hardware. Your model. Your machine. Not every decision needs the cloud.

FLock.io

12,737 次观看 • 1 个月前

Introducing Vela 3.0 📡 Name it. Vela finds it. Six open models (0.3B–27B): decisions + the spans you name. One call. ⚡ 380M: #1 Decision Index ≤500M 🎯 Beats GPT-5-mini at zero-shot classification 🔍 18 ms per span question 🤗 #Jev #Decison

Introducing Vela 3.0 📡 Name it. Vela finds it. Six open models (0.3B–27B): decisions + the spans you name. One call. ⚡ 380M: #1 Decision Index ≤500M 🎯 Beats GPT-5-mini at zero-shot classification 🔍 18 ms per span question 🤗 #Jev #Decison

Xunzhuo

26,647 次观看 • 18 天前

Introducing JEV-27B-VL. We believe it is the world’s first open-weight, near-SOTA multimodal decision model — built with AutoTrust AI’s Blocks of Experts (BoE) recipe. It does not just describe what it sees. It turns visual state into calibrated action probabilities, then selects what to do next. In this demo, it sees Mario’s position, movement, obstacles, and timing — then chooses whether to run, jump, or do both. See → decide → act.

Introducing JEV-27B-VL. We believe it is the world’s first open-weight, near-SOTA multimodal decision model — built with AutoTrust AI’s Blocks of Experts (BoE) recipe. It does not just describe what it sees. It turns visual state into calibrated action probabilities, then selects what to do next. In this demo, it sees Mario’s position, movement, obstacles, and timing — then chooses whether to run, jump, or do both. See → decide → act.

AutoTrust

155,683 次观看 • 1 个月前

updated Decision Index 0.1 → 0.2 🎯 better formula, +29 jev-like models, +21 benchmarks AutoJev-27B by Perplexity CTO Denis Yarats took the open lead, trailing jev by 0.8 points 🏆 come find the best model at every size, speed and use-case

updated Decision Index 0.1 → 0.2 🎯 better formula, +29 jev-like models, +21 benchmarks AutoJev-27B by Perplexity CTO Denis Yarats took the open lead, trailing jev by 0.8 points 🏆 come find the best model at every size, speed and use-case

apolinario (poli)

24,255 次观看 • 1 个月前

Drex 1.5 is Live! 128k context window, more rivalry to Jev #1 on the Decision Index 0.2.1 + JevBench (SOTA) Absolute leader in long context decisions Latency: sub-second Decision models are getting better! Price: $0.04 per 1M input tokens Try Drex with 250M free tokens #drex #jev #nace

Drex 1.5 is Live! 128k context window, more rivalry to Jev #1 on the Decision Index 0.2.1 + JevBench (SOTA) Absolute leader in long context decisions Latency: sub-second Decision models are getting better! Price: $0.04 per 1M input tokens Try Drex with 250M free tokens #drex #jev #nace

Nace AI

541,219 次观看 • 1 个月前

Stop using your most expensive model for every decision. In our Minesweeper experiment (open-source code below), JEV crushed every other model we tested on speed and cost. JEV: 8s / ~$0.003 Fable 5.1: 100s / ~$2.41 Astra: 128s / ~$8.75 Grok 4.6: 158s / ~$0.55 Where would you put JEV to work in your company?

Stop using your most expensive model for every decision. In our Minesweeper experiment (open-source code below), JEV crushed every other model we tested on speed and cost. JEV: 8s / ~$0.003 Fable 5.1: 100s / ~$2.41 Astra: 128s / ~$8.75 Grok 4.6: 158s / ~$0.55 Where would you put JEV to work in your company?

Buddy Works

171,555 次观看 • 1 个月前

Introducing Mercury Decide: the most intelligent decision model on OpenRouter (JevBench v1.4) and one of the fastest. Up to 14 decisions per second. Watch it go head-to-head with Jev in chess. Free for early access on OpenRouter:

Introducing Mercury Decide: the most intelligent decision model on OpenRouter (JevBench v1.4) and one of the fastest. Up to 14 decisions per second. Watch it go head-to-head with Jev in chess. Free for early access on OpenRouter:

Inception

20,338 次观看 • 1 个月前

Decision 3.0 now on Core ML 🍎 d3-lite triaging 1,000 pull requests into vllm-sr workgroups, live on an M5 Pro: 0.12 s per PR, 8.3 PRs/s, nothing leaves the Mac ⚡ lite, nano and mini are all converted and in one repo 👇 🤗 🛠️

Decision 3.0 now on Core ML 🍎 d3-lite triaging 1,000 pull requests into vllm-sr workgroups, live on an M5 Pro: 0.12 s per PR, 8.3 PRs/s, nothing leaves the Mac ⚡ lite, nano and mini are all converted and in one repo 👇 🤗 🛠️

Alex

28,713 次观看 • 1 个月前