# Decision 2.0 and Vela 2.0 vs Kev-4B: benchmark on a GB10 > vLLM Semantic Router has released two families of open decision models, Decision 2.0 and Vela 2.0. On a GB10, Vela 2.0 4B named the kind of task right on 96.6% of our 328 requests against 92.4% for Kev-4B, at three times the wait, and its 0.3B sibling reached 85.7% without a GPU. - Published: 2026-10-07 - Author: Flavien Berwick - Category: Research - Web page: https://carouter.ai/blog/decision-2-and-vela-2-vs-kev - In French: https://carouter.ai/fr/blog/decision-2-and-vela-2-vs-kev.md ![Research: Decision 2.0 and Vela 2.0 vs Kev-4B, benchmark on a GB10. Kind of task named right on our 328 requests: Decision 2.0 Nox-4B 93.0% in 194 ms and 11 GB, Vela 2.0 4B 96.6% in 319 ms and 23 GB, Kev-4B 92.4% in 99 ms and 14 GB, Vela 2.0 0.3B without a GPU 85.7% in 92 ms and 4 GB. OpenJev's untrained Qwen3.5-4B still leads at 97.0%, in 192 ms and 9 GB.](https://carouter.ai/blog/decision-2-and-vela-2-vs-kev/illustration.png) The vLLM Semantic Router project has released two families of open decision models: Decision 2.0, and this week Vela 2.0. Both answer typed questions about a text the way Jev and Kev do. We ran three of them beside Kev on our two benchmarks, on an NVIDIA GB10, the machine that serves Kev-4B at /v1/decisions. They are Decision 2.0 Nox-4B, Vela 2.0 4B, and Vela 2.0 0.3B, which is small enough to run without a GPU. This is the fifth article of the series. Its table holds every decision model we have tested so far, and its last sections say what each one is good for. [Read the first benchmark.](https://carouter.ai/blog/jev-kev-laya-decision-models) [Read the second, with Kev's launch.](https://carouter.ai/blog/kev-available-on-carouter) [Read the third, on OpenJev.](https://carouter.ai/blog/openjev-vs-kev) [Read the fourth, on Clef and Strands Decider.](https://carouter.ai/blog/clef-strands-decider-vs-kev) ## The short version - **On our requests.** Vela 2.0 4B named the kind of task right on 96.6% of our 328 requests, against 93.0% for Decision 2.0 Nox-4B and 92.4% for Kev-4B. It is the most accurate trained model we have tested, one request behind OpenJev's untrained Qwen3.5-4B. - **Where Vela gains.** It recognised every casual chat and 29 of the 32 agent steps. CARouter reads an agent step from the request itself, and scored that way Vela 2.0 4B is at 97.6% and Kev-4B at 97.3%, one request apart. - **Without a GPU.** Vela 2.0 0.3B reached 85.7% on a processor alone, ahead of Kev-0.8B on a GPU at 80.2%. A short decision took 92 milliseconds. - **On public prompts.** Vela 2.0 4B reaches 80% in English and 76% in French, and Nox-4B 79% and 78%. Kev-4B is at 76% and 70%. - **Confidence.** On public prompts, Nox-4B was wrong while at least 90% sure on 0.7% of the English requests, against 3% for Kev-4B and 7% for Vela 2.0 4B. - **Speed.** A three-question decision on a short request takes 99 milliseconds on Kev-4B, 194 on Nox-4B and 319 on Vela 2.0 4B. - **Memory.** Vela 2.0 4B holds 23 GB of the GB10's 121, Kev-4B 14 and Nox-4B 11. The 0.3B holds 1.5 GB on the GPU. ## What the newcomers are Decision 2.0 is a family of six models, from 0.6 to 27 billion parameters. We ran its 4B model, Nox, released under the Apache 2.0 licence. It is a fine-tune of Qwen3.5-4B-Base, the model Kev-4B is built on, and it answers the three kinds of question Jev does: a choice, a yes or no, and a score. [Nox-4B's weights.](https://huggingface.co/vllm-sr/Decision-2.0-Nox-4B) Vela 2.0's larger models start from Decision 2.0. They are trained for what a router must know about a request before sending it anywhere: its subject, whether it is an attack, whether it carries personal information, whether an answer is backed by its sources. [The announcement.](https://vllm-sr.ai/blog/vela-2-0-open-foundation-routing-models/) [Vela 2.0 4B's weights.](https://huggingface.co/vllm-sr/Vela-2.0-4B) It adds two kinds of answer. A set question returns every label that applies, and a span question returns the exact words it found, with their position in the text. Vela 2.0 0.3B is built differently. It is an encoder of 0.3 billion parameters, the kind of model that reads a text without being able to write one, and it runs on a processor. Its weights are under Apache 2.0, and its tokenizer keeps Google's Gemma terms of use. Their authors say the specialisation has a price. On the Jev Decision Index, a public suite of general decision tasks, the model cards report about 32 for Vela 2.0 4B and about 44 for the Nox-4B it started from. We expected Vela to read our requests worse than Nox, and the opposite happened. The two 4B models read up to 16,384 tokens and the 0.3B reads 8,192, questions included. Decision 2.0 refuses a request that is too long, and Vela 2.0 cuts it. None of our test requests reached those limits. ## How we tested The test machine is a GB10, the same model as the one that serves Kev, with 121 GB of memory shared between the processor and the GPU. Each model ran alone on it, one request at a time, through its authors' own code. Kev ran the way our server runs it. The two benchmarks are those of the earlier articles. On the 1,020 public prompts, each model answers one question with nine kinds of task. On our 328 requests it answers three: the kind of task, how hard it is, and whether the answer will be long. One script scored every model, the one of the previous article. Before scoring anything new, it recomputes the published figures of the earlier articles and stops if one differs. The models of the earlier articles keep the figures we published. Kev-4B ran again on this test machine as a reference. It chose the same answer as last time for 1,016 of the 1,020 public prompts and 325 of our 328 requests, and its timings are within 2% of the last article's. Both Vela models gave their model card's answers on the card's own example, to within 0.001. Decision 2.0 publishes no answer to compare with. It checks every one of its files against a manifest before loading, and it answered its card's example sensibly. The 0.3B ran three ways: on the GPU, and on the processor alone through two inference engines. All three chose the same answer on every request, so its accuracy has one row and its speed has two. Jev is a paid hosted service that we did not call again. Its figures on the public prompts are those of the first benchmark, and it has never been run on our requests. ## What we found: accuracy ![Kind of task named right, in English and in French, for every model in the table. On the 1,020 public prompts: Jev 81% and 76%, Kev-9B 74% and 67%, Kev-4B 76% and 70%, Kev-0.8B 70% and 64%, OpenJev Qwen3.5-4B 79% and 81%, Strands Decider 80% and 68%, Clef-flash 82% and 78%, Clef 82% and 76%, Decision 2.0 Nox-4B 79% and 78%, Vela 2.0 4B 80% and 76%, Vela 2.0 0.3B 71% and 64%. On our 328 requests: Kev-4B 93% and 92%, Kev-0.8B 80% and 81%, OpenJev Qwen3.5-4B 97% and 97%, Strands Decider 89% and 85%, Clef-flash 90% and 90%, Clef 95% and 94%, Decision 2.0 Nox-4B 93% and 93%, Vela 2.0 4B 96% and 97%, Vela 2.0 0.3B 87% and 85%. Jev and Kev-9B were not run on our requests.](https://carouter.ai/blog/decision-2-and-vela-2-vs-kev/01-accuracy-both-benchmarks.png) Vela 2.0 4B got 317 of our 328 requests right. OpenJev's Qwen3.5-4B got 318, Clef 310, Nox-4B 305 and Kev-4B 303. We wrote the questions about our requests with Kev-4B's answers in front of us, on three fifths of the requests. On the 137 requests kept aside, Vela 2.0 4B missed one: 99.3%, against 98.5% for OpenJev, 94.9% for Nox-4B and 93.4% for Kev-4B. On the public prompts the newcomers sit with the best of the earlier models. Vela 2.0 4B reaches 80% in English and Nox-4B 79%, around the hosted Jev at 81% and ahead of Kev-4B at 76%. French costs Nox-4B one point there, 79% against 78%. That is the smallest drop of any trained model in the table. Vela 2.0 4B loses five points, like Jev, and Kev-4B six. **Every decision model we have tested, on both benchmarks and on the GB10** | Model | Public prompts, English | Public prompts, French | Confidently wrong, public, English | Our 328 requests | Our 328 requests, agent steps read from the request | Short decision | Memory | | --- | --- | --- | --- | --- | --- | --- | --- | | Jev (hosted) | 80.9% | 76.0% | 11.7% | not run | not run | not run | not run | | Kev-9B | 74.3% | 66.5% | 4.8% | not run | not run | not run | not run | | Kev-4B | 75.7% | 69.8% | 3.3% | 92.4% | 97.3% | 99 ms | 14 GB | | Kev-0.8B | 69.6% | 63.7% | 0.0% | 80.2% | 86.3% | 23 ms | 4 GB | | OpenJev Qwen3.5-4B | 79.3% | 80.8% | 19.3% | 97.0% | 97.6% | 192 ms | 9 GB | | Strands Decider 2B | 80.4% | 67.7% | 0.2% | 87.2% | 91.2% | 108 ms | 4 GB | | Clef-flash (9B) | 82.0% | 77.9% | 8.9% | 89.6% | 95.1% | 173 ms | 18 GB | | Clef (27B) | 81.9% | 75.8% | 8.1% | 94.5% | 95.7% | 562 ms | 52 GB | | Decision 2.0 Nox-4B | 78.9% | 77.7% | 0.7% | 93.0% | 96.6% | 194 ms | 11 GB | | Vela 2.0 4B | 80.4% | 75.6% | 6.7% | 96.6% | 97.6% | 319 ms | 23 GB | | Vela 2.0 0.3B, on CPU | 71.1% | 64.0% | 3.3% | 85.7% | 88.4% | 92 ms | 4 GB | Vela 2.0 0.3B got 281 of our requests right, 85.7%. That is 18 more than Kev-0.8B and 5 fewer than Strands Decider, two models that need a GPU. On the public prompts it is level with Kev-0.8B, at 71% in English and 64% in French. ## Where the differences come from ![Kind of task named right on our 328 requests, by kind, for Kev-4B, OpenJev Qwen3.5-4B, Clef, Decision 2.0 Nox-4B, Vela 2.0 4B and Vela 2.0 0.3B. Chat: 87%, 100%, 83%, 93%, 100% and 97%. Agent step: 50%, 94%, 88%, 63%, 91% and 72%. Research: 82%, 75%, 71%, 75%, 71% and 21%. Summarization: 100% for all but the 0.3B at 85%. Extraction: 100%, 97%, 97%, 94%, 100% and 94%. Math and business writing: 100% for all but the 0.3B at 91% and 86%. All six reach 100% on code and translation.](https://carouter.ai/blog/decision-2-and-vela-2-vs-kev/03-by-kind.png) Vela 2.0 4B is perfect on seven of the nine kinds. It recognised all 30 casual chats, where Nox-4B recognised 28 and Kev-4B 26. Agent steps are where it gains most. It recognised 29 of 32, against 20 for Nox-4B and 16 for Kev-4B. The latest message of an agent step is a tool's result, which reads like code or a web page. It is also the one kind CARouter does not ask a model about. A request that carries tool definitions is an agent step, and our platform sees that the way it sees the language. Scored that way, Vela 2.0 4B names 97.6% of the requests, Kev-4B 97.3% and Nox-4B 96.6%. Research is the one kind where Kev-4B leads the newcomers: 23 of 28, against 21 for Nox-4B and 20 for Vela 2.0 4B. The 0.3B recognised 6 and filed 14 of the others under chat. ## What we found: confidence A model that decides on someone's behalf should act only when it is sure. So we count how often each one is wrong while at least 90% sure. ![Confidence on both benchmarks. On the public prompts in English, accuracy against wrong answers given at 90% confidence or more: Jev 81% right and 12% confidently wrong, Kev-9B 74% and 5%, Kev-4B 76% and 3%, Kev-0.8B 70% and 0%, OpenJev Qwen3.5-4B 79% and 19%, Strands Decider 80% and 0.2%, Clef-flash 82% and 9%, Clef 82% and 8%, Decision 2.0 Nox-4B 79% and 0.7%, Vela 2.0 4B 80% and 7%, Vela 2.0 0.3B 71% and 3%. On our 328 requests, answers given at 90% confidence or more: Kev-4B 187 with none wrong, Kev-0.8B 2 with none wrong, OpenJev Qwen3.5-4B 292 with 4 wrong, Strands Decider 131 with none wrong, Clef-flash 227 with 2 wrong, Clef 213 with none wrong, Decision 2.0 Nox-4B 59 with none wrong, Vela 2.0 4B 269 with 2 wrong, Vela 2.0 0.3B 172 with none wrong.](https://carouter.ai/blog/decision-2-and-vela-2-vs-kev/02-confidence.png) On the public prompts, Nox-4B is confidently wrong on 0.7% of the English requests and on none of the French ones. Vela 2.0 4B is at 7%, between Kev-4B at 3% and Clef at 8%. On our requests, Vela 2.0 4B gave 269 answers at that level and two were wrong. The 0.3B gave 172, Kev-4B 187 and Nox-4B 59, with none wrong. Each model computes its confidence its own way, and Decision 2.0's is the strictest: it falls as soon as a second option gets some probability. Recomputed with Kev's formula, Nox-4B is that sure of 144 answers, and its confident mistakes on the English public prompts rise from 0.7% to 5%. ## Difficulty and long answers Both 4B newcomers rate difficulty better than Kev-4B. Nox-4B places a request within one level of our rating 94% of the time and Vela 2.0 4B 92%, against 88% for Kev-4B and 81% for the 0.3B. Clef still leads at 98%. Ten of our 328 requests need a long answer. Nox-4B flagged all ten and five others. Vela 2.0 4B found eight and flagged eight others, and Kev-4B found seven and flagged no other. The 0.3B cannot answer that question. It found four of the ten and flagged 154 requests that did not need a long answer. ## Speed and memory on a GB10 We timed the three-question decision of the earlier articles on a new request each time, from a short request to a 4,000-token one. ![Median time of a three-question decision by prompt length on the GB10, on log scales, and memory held at its peak. Kev-4B: 99 to 631 ms, 14.4 GB. Decision 2.0 Nox-4B: 194 ms to 3.1 s, and 1.1 s on the longest prompt with its shared-context option, 11.2 GB. Vela 2.0 4B: 319 ms to 1.3 s, 22.8 GB. Vela 2.0 0.3B on the GPU: 11 to 243 ms, 1.5 GB. Vela 2.0 0.3B on the processor alone: 92 ms to 3.9 s, 3.8 GB.](https://carouter.ai/blog/decision-2-and-vela-2-vs-kev/04-latency-memory-gb10.png) Kev-4B stays the fastest of the 4B models, at 99 milliseconds on a short request and 631 on a long one. Vela 2.0 4B takes 319 milliseconds and 1.3 seconds. Nox-4B takes 194 milliseconds and 3.1 seconds. Nox-4B is slow on long requests because it reads the request again for every question. It has an option that reads it once, which brings the long request down to 1.1 seconds and changed none of its answers on our two benchmarks. Asked a single question, Nox-4B is the fastest 4B model here: 72 milliseconds on a short request, against 83 for Kev-4B. The 0.3B is in another class. On the GPU it takes 11 milliseconds on a short request and 243 on a long one, in 1.5 GB. On the processor alone, using ten of its twenty cores, it takes 92 milliseconds on a short request, 581 on a 1,000-token one and 3.9 seconds on a 4,000-token one. In our third article, the fastest model on a GB10's processor was Kev-0.8B, at 1.3 seconds on a short request, with the cores shared with other work. Vela 2.0 4B holds 23 GB, because its authors' code keeps its weights as 32-bit numbers. Kev-4B holds 14 GB and Nox-4B 11. On a processor, the 0.3B holds 4 GB of ordinary memory. These timings are the best case for every model: alone on the machine, with no network in between. In production, Kev-4B shares its GB10 with other models and answers through an API, and the same decision takes 143 milliseconds and 1 second. ## What each model is good for A decision model answers a closed question about a text and says how sure it is. After five articles, here is where each model of the table fits. - **Choosing a model for each request.** This is what our 328 requests measure. OpenJev's Qwen3.5-4B, Vela 2.0 4B and Clef read them best. Kev-4B is one request behind Vela once agent steps are read from the request, and it answers three times sooner. - **Decisions inside your own application.** This covers sorting support tickets, sending a message to the right team, checking a condition before an action and rating urgency. Kev-4B does it today at /v1/decisions. Nox-4B fits where French matters or where a confident mistake is the costly one. - **Decisions without a GPU.** Vela 2.0 0.3B answers on a processor while a caller waits, as long as the request is short. It is reliable on clear-cut kinds such as code, translation and extraction, and not on research or on long answers. Laya, in our first article, also answered in a fraction of a second on a processor, and it named 58% of the English public prompts right, against 71% here. - **Decisions beside a large model.** When the GPU's memory belongs to a language model, the small ones matter: the 0.3B in 1.5 GB, Kev-0.8B and Strands Decider in 4 GB. - **Checking a request before it leaves.** Vela 2.0 was trained to find personal information, prompt attacks, unsafe content and unsupported claims in the same read as its other questions. We tested none of that here. On CARouter, the PII filter and the content filter do this work with two separate models. - **Images and video.** Clef is the only model here that reads them, and we tested text only. - **Nothing to run.** Jev is a hosted service: there is no model to serve, and every request leaves your infrastructure. ## What this does not prove - This is one run per model on two sets of requests. A kind holds 20 to 66 of our requests, and on the smallest, one mistake moves the figure by five points. - We wrote and labelled the 328 requests, and we tuned their questions on Kev-4B. Both benchmarks ask one thing: what kind of work a request is. On general decision tasks, Vela 2.0's own authors measure it below Decision 2.0. - We did not test what Vela 2.0 was built for. Nothing here says how well it finds personal information, prompt attacks or unsupported claims, in English or in French. - The new models ran through the code their authors published. Decision 2.0's fastest path is written for AMD GPUs and did not apply on the GB10's NVIDIA GPU, and Kev's serving code is tuned for this machine. The time gaps say as much about the software as about the models. - The 0.3B was timed on the GB10's processor, an Arm chip with twenty cores. Another processor will give other timings. - Jev was not called again. Its figures come from the first benchmark, and it was never run on our requests. ## What this means for CARouter Kev-4B stays behind /v1/decisions for now. Scored the way our platform reads a request, Vela 2.0 4B is one request ahead of it, for three times the wait and 8 GB more memory on a machine that serves other models. Vela 2.0 4B is still the result of this article. It is the first trained model to read our requests about as well as the Qwen3.5-4B nobody trained, and it could answer routing questions and safety questions in one read. Before that changes anything, we have to test the safety half ourselves, in both languages. Nox-4B is two requests ahead of Kev-4B, steadier in French and better at difficulty and long answers. It is also twice as slow on our three questions, which is not enough to switch. The 0.3B changes one assumption. In our third article we concluded that a processor was no place for Kev or OpenJev while a caller waits. The 0.3B answers there in a tenth of a second on a short request, and it is a candidate wherever a deployment has no GPU to spare. Every measurement behind this article is kept, and we will share it with anyone who asks. --- More articles: [CARouter Blog](https://carouter.ai/blog.md). What CARouter is and how to call it: [llms.txt](https://carouter.ai/llms.txt).