“Zero output tokens” means the model does not decode a text answer. It still processes the input in a forward pass, then reads internal hidden states at designated answer positions to choose among options the caller has already specified. The phrase describes how the answer is produced—not how much computation or input the model handles.
How can a model answer without generating tokens?
The caller provides a state and one or more questions, each paired with an ordered set of allowed answers. For example, an application might ask whether a condition is true or false, or ask the model to choose one of several named actions. The rendered request includes designated positions associated with those answers.
- The model processes the rendered request in a forward pass.
- It reads hidden states at the designated answer positions.
- A softmax over the declared options produces a probability distribution, from which the application gets a typed result.
Because the output head is limited to the declared options, it cannot return an answer outside that set. The model does not sample or decode an open-ended text response for application code to interpret. The paper says multiple questions about one state can be handled in a single forward pass. Zehua Cheng, Wei Dai, and Jiahao Sun describe this mechanism in their 2026 paper, “this-that-model-1.0: A typed decision model that decides in 30 ms, for a millionth of a cent”.
Does zero output tokens mean the model did no computation?
No. The model still processes the input and performs the computations needed to produce hidden states and option probabilities. “Zero” refers only to decoded output tokens: there is no generated textual answer. It does not mean zero input tokens, zero processing, or an instant response.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →#1 Best Overall
- EVOLUTION AMD RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 128GB pool, which is perfect for running LLMs such as Deepseek 70B Q8, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 12% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
Why use a typed decision instead of generated text?
For software decisions with a known set of valid answers, the paper’s approach replaces a generate-then-parse workflow with a constrained result. That can avoid decoding a text string and handling parser errors or answers that are malformed or missing. It also gives the application probabilities over the allowed options. This is useful when the application can define the answer set in advance and send uncertain cases to another process.
The trade-off is that the model answers the bounded question it was given; this interface is not a general-purpose text response. Cheng, Dai, and Sun report 30.9 milliseconds per decision and 32 decisions per second on one consumer GPU in their setup. Those are measurements from the paper, not guarantees for different hardware, software stacks, request shapes, or production conditions. The authors also compare against hosted-model measurements, but the configurations and cost bases differ.
Rank #2
- Built for Local AI Development: AMD Ryzen AI Halo is designed for local AI development and inference, featuring 128GB unified memory and support for up to 200B parameter models to build and run intensive AI workloads locally.
- 128GB Unified Memory: Features 128GB LPDDR5x unified memory at 8000 MT/s with 256 GB/s memory bandwidth, providing a shared memory pool across the CPU, GPU, and NPU to support larger AI models.
- AMD Ryzen AI Max+ 395 Processor: Features 16 cores, 32 threads, and Zen 5 architecture, paired with AMD Radeon 8060S integrated graphics featuring 40 RDNA 3.5 compute units and an AMD XDNA 2 NPU with up to 50 TOPS.
- Linux AI Developer Platform: Purpose-built for Linux-based AI development with full AMD ROCm software support and preloaded tools, models, and workflows optimized for local AI development.
- Compact, Connected Design: Includes a 2TB M.2 SSD, 10GbE LAN, Wi-Fi 7, Bluetooth 5.4, USB-C connectivity, and HDMI 2.1b.
What does the paper’s evidence establish—and what does it not?
The paper evaluates a specific software model on particular decision tasks. Its figures should be read as results for those setups rather than as a general claim about multimodal models or hosted systems.
A small third-party comparison
On a third-party recorded cohort of 68 decision questions, the authors report accuracy of 0.941 and a Brier score of 0.042 for this-that-model-1.0. Jev scored 0.765 accuracy and 0.133 Brier score on the same items. The authors note that the cohort is small, its wording came from the third party, and the accuracy gap rests on 12 questions. This is not broad proof that the model is more accurate than hosted frontier models.
Rank #3
- EVOLUTION RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 128GB pool, which is perfect for running LLMs such as Deepseek 70B Q8, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 12% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
Arithmetic and search are weak spots
On multi-step arithmetic, the model scores 0.560, compared with 0.98 to 1.00 for the hosted systems cited in the paper. The authors attribute this limitation to the single forward pass, which cannot carry intermediate results through a sequence of calculations. Their released benchmark contains 7,305 questions across 15 families and two environments; results vary by task, and map-wide search questions remain a persistent weakness. Their conclusion is that direct mappings are a better fit than tasks that require the model to perform a search.
Probability quality depends on the task
On one constructed stochastic-actuator evaluation, the paper reports a score of 0.750 against an estimated ceiling of 0.746. That result applies to that evaluation; it is not a universal guarantee of calibration or probability quality.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How does this differ from other decision interfaces?
| Approach | Output contract | Best fit | What to check |
|---|---|---|---|
| This-that-model-1.0 | Probabilities over options declared by the caller; no decoded text answer. | Bounded decisions that can be learned as a direct mapping. | Whether the task resembles the evaluations, and whether local execution and the model’s probability output suit the application. |
| Generative model | Generated text that application code may need to parse. | Requests that need an open-ended response; reliability depends on the model and task. | How the application validates the response and handles malformed, missing, or out-of-set answers. |
| Hosted typed-decision service | Depends on the service’s interface; may return a constrained result. | Depends on the service and its supported tasks. | Compare its output contract, latency, probability access, deployment and data handling, and evaluation evidence with the actual use case. |
The paper describes this-that-model-1.0 as an open-source software model and inference code, not a physical product. Its “multimodal” title should also be interpreted narrowly: the described request can be a string or compactly serialized JSON, while the examples and reported benchmarks focus on structured decisions and map-like environments. The paper does not establish performance across every image, audio, or video task. Claims about local execution should not be treated as independent verification of a particular deployment’s operational or data-handling behavior.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →
