01ESCALATING2 signals
Embodied generalisation
Can orchestration and compositional policies survive unfamiliar tasks, environments and robot bodies?
Singularity PressureCurrent index
Moderate
Live pressure3.5/ 10
System stateAccelerating↑
A qualitative index built from recent signal intensity. It is not a probability or a prediction of when a singularity will happen. Real published proposals only.
3.5/ 10
010System state
- Stable0—2.4
- Accelerating2.5—4.4
- Unstable4.5—6.4
- Breakthrough6.5—8.4
- Runaway8.5—10
Several frontier lines moved at once. The evidence is real enough to watch, not strong enough to call a break.
Weekly synthesis
This week moved the needle because…
01
Autonomy jump
+2.00 index pts02Robotics moved toward video-grounded task orchestration and multi-robot collaboration—not general autonomy, but a broader control surface.
Autonomy jump
+0.78 index pts03Agentic scores rose sharply across terminal, software-engineering and automation evaluations. The results remain publisher-reported.
Capability delta
+0.43 index ptsPreserving reasoning state and compacting context tripled performance on ARC-AGI-3. A narrow result, with an important mechanism.
Active vectors / 05
What we're watching
Not predictions. The live questions behind the current index—and the evidence needed to move beyond a promising signal.
02WATCHING1 signal
Agentic endurance
Do evaluation gains become reliable long-horizon execution outside publisher-controlled tests?
03WATCHING1 signal
Reasoning persistence
Does retained reasoning state transfer beyond one benchmark into harder, less structured problem spaces?
04UNCONFIRMED1 signal
The serving bottleneck
Do stateful infrastructure gains materially lower latency and cost once agents operate at scale?
05UNCONFIRMED1 signal
Control boundaries
Can evaluation environments reliably prevent capable models from reaching unauthorised real systems?
Published record
Signal log
Published assessments, ordered as frontier signals rather than news stories.
6 signals
Google DeepMind · 03 AUG
Autonomy jumpGemini Robotics ER 2: powering robotics with video understanding, task orchestration, and multi-robot collaboration
Gemini Robotics ER 2 helps robots reason, collaborate, and solve real-world tasks. It represents a step change in video understanding, tool orchestration, and multi-robot collaboration for robotic applications.
- Reality
- 6.5
- Singularity pressure
- 5.2/10
- Confidence
- 76%
OpenAI · 03 AUG
Capability deltaHow enabling two settings tripled our scores on the ARC-AGI-3 benchmark
How two API settings improved GPT-5.6 performance on ARC-AGI-3, boosting scores and efficiency by retaining reasoning and enabling compaction.
- Reality
- 6.5
- Singularity pressure
- 3.7/10
- Confidence
- 76%
Anthropic · 03 AUG
Control signalInvestigating three real-world incidents in our cybersecurity evaluations
In a review of our cybersecurity evaluation transcripts, we found three incidents in which a Claude model reached the internet from within or while interacting with a third-party evaluation environment, and then gained unauthorized access to the real systems of three different organizations. Below we describe what happened, how it happened, and what we’re changing. We encourage other AI labs to perform similar reviews.
- Reality
- 3.0
- Singularity pressure
- 1.2/10
- Confidence
- 68%
arXiv · 03 AUG
Autonomy jumpDiagnosing Compositional Generalization in Sequential Robot Tasks
Sequential robot manipulation requires policies to execute novel combinations of familiar instruction components. However, collecting demonstrations for all possible instruction tuples is combinatorially expensive, while sparsely covered datasets often fail under out-of-distribution recombination. This paper studies compositional generalization through the lens of instruction-space coverage. We decompose the generalization gap into three sources: \textit{marginal instruction shift}, \textit{instruction-compositional shift}, and \textit{context--action shift}. This decomposition allows us to diagnose when sparse training coverage is sufficient, and what structure the training set must preserve for reliable action prediction. Our results show that exhaustive tuple enumeration is unnecessary: a structured subset, as small as one quarter of the full task space, can recover strong out-of-distribution performance when it covers action-relevant dependencies. We further find that sparse training often fails due to instruction steering rather than missing low-level skills; finetuning only one demonstration per task improves OOD success from \(0.4\%\) to \(54.7\%\). For semantically dependent tasks, effective coverage must capture relational structure rather than only factor diversity. These findings suggest that efficient robot data collection should prioritize dependency coverage in instruction space over exhaustive task expansion. More results are available in the supplementary material. Project website: https://yixiaowang7.github.io/Diagnosing_Compositional_Generalization_Robot_Page/.
- Reality
- 6.0
- Singularity pressure
- 3.2/10
- Confidence
- 68%
arXiv · 03 AUG
Compute shockTokTier: Exact Stateful Tokenization for Agentic LLM Serving
LLM serving systems cache prompt KV state, yet most front ends still re-tokenize the full request text on every call. The cost lands on coding agents, which resubmit a long transcript after each small tool result, and reuse is hard because even a short append can change token boundaries near the end of the previous sequence. Across 153,951 calls from two agent ecosystems, the median call appends about 1.4K characters, and only 1.0-3.6% of calls start or rebuild a session with contexts of millions of characters. At a 94.1% fleet prompt-cache hit rate, tokenization reaches up to 64% of time to first token. TokTier is a stateful tokenization service with one contract: emitted token IDs are always identical to full reference tokenization of the request text. For a session continuation, it re-tokenizes a small window around the append and splices only after a per-request stable-boundary check, widening the window or falling back to full tokenization on failure. For a call without a reusable prefix, it decomposes GPT-family regex pre-tokenization into run-local rules and runs exact pre-tokenization and BPE on a GPU. A sampled shadow verifier re-checks live traffic. Across 17 tokenizer families, differential campaigns cover 1.5x10^10 split checks, a 12.4 TB real-text corpus, and 93,000+ replayed agent steps, with zero divergence. Incremental repair takes 0.5-1.1 ms from 100K to 3M characters, up to 437x faster than HF tokenization and 2.1x faster at 1M than the strongest cache-based baseline (Gigatoken) fully prewarmed. GPU full tokenization encodes a 1M-character request in 0.87 ms, up to 491x below HF and 23.4x below the fastest published CPU method. With vLLM, median time to first token drops 16-34% and P99 drops 23% under recorded bursts. Under a 50 ms P99 objective, four repair cores plus one GPU sustain 1,821 requests/s where a 16-core stateless front end saturates at 40.
- Reality
- 3.0
- Singularity pressure
- 2.2/10
- Confidence
- 68%
Hugging Face · 03 AUG
Autonomy jumpdeepseek-ai/DeepSeek-V4-Flash-0731
{"author":"deepseek-ai","pipelineTag":"text-generation","tags":["transformers","safetensors","deepseek_v4","text-generation","conversational","arxiv:2606.19348","license:mit","eval-results","endpoints_compatible","8-bit","fp8","region:us"],"cardData":{"license":"mit","library_name":"transformers"}} --- license: mit library_name: transformers --- # DeepSeek-V4-Flash-0731 Technical Report 👁️ ## Introduction **DeepSeek-V4-Flash-0731** is the official release of **DeepSeek-V4-Flash**, superseding the preview version, with substantially enhanced agentic capabilities. It has the same model structure as [DeepSeek-V4-Flash-DSpark](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-DSpark), i.e. it comes with a speculative decoding module attached. DeepSeek-V4-Flash-0731 outperforms DeepSeek-V4-Pro (Preview) on benchmarks listed below despite its far smaller activated parameter count, and is broadly competitive with the strongest proprietary models available. | Benchmark | DeepSeek-V4-Flash-0731 | DeepSeek-V4-Flash (Preview) | DeepSeek-V4-Pro (Preview) | GLM-5.2 | Opus-4.8 | | :--- | :---: | :---: | :---: | :---: | :---: | | Terminal Bench 2.1 | 82.7 | 61.8 | 72.1 | 81.0 | 85.0 | | NL2Repo | 54.2 | 39.4 | 38.5 | 48.9 | 69.7 | | Cybergym | 76.7 | 38.7 | 52.7 | - | 83.1 | | DeepSWE | 54.4 | 7.3 | 12.8 | 46.2 | 58.0 | | Toolathlon-Verified | 70.3 | 49.7 | 55.9 | 59.9 | 76.2 | | Agents' Last Exam | 25.2 | 15.8 | 16.5 | 23.8 | 25.7 | | AutomationBench Public | 25.1 | 10.8 | 12.8 | 12.9 | 27.2 | | DSBench-FullStack † | 68.7 | 37.0 | 41.8 | 61.8 | 71.6 | | DSBench-Hard † | 59.6 | 25.8 | 31.1 | 54.5 | 71.7 | Notes: 1. For the Code Agent tasks among the public benchmarks above, DeepSeek-V4-Flash-0731 is evaluated with the minimal mode of DeepSeek Harness (to be released) as the agent framework, using the `max` reasoning effort level with `temperature = 1.0, top_p = 0.95`. 2. † DSBench-FullStack is an internal full-stack development test set; DSBench-Hard is an internal test set of difficult coding-agent problems. ## Chat Template This release does not include a Jinja-format chat template. Instead, we provide a dedicated `encoding` folder with Python scripts and test cases demonstrating how to encode messages in OpenAI-compatible format into input strings for the model, and how to parse the model's text output. Please refer to the [`encoding`](encoding/README.md) folder for full documentation. The `reasoning_effort` parameter now supports three levels — `low`, `high`, and `max` — which control how much deliberation the model spends before answering. A brief example: ```python from encoding_dsv4 import encode_messages, parse_message_from_completion_text messages = [ {"role": "user", "content": "hello"}, {"role": "assistant", "content": "Hello! I am DeepSeek.", "reasoning_content": "thinking..."}, {"role": "user", "content": "1+1=?"} ] # messages -> string prompt = encode_messages(messages, thinking_mode="thinking", reasoning_effort="max") # string -> tokens import transformers tokenizer = transformers.AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-V4-Flash-0731") tokens = tokenizer.encode(prompt) ``` ## How to Run with vLLM DSpark speculative decoding is enabled with a single flag — add --speculative-config with method: dspark to your vLLM launch command: `--speculative-config '{"method":"dspark","num_speculative_tokens":7,"draft_sample_method":"greedy"}'` For example, the command below serves the model with vLLM on a single 4×GB300 node. See the [vLLM recipe](https://recipes.vllm.ai/deepseek-ai/DeepSeek-V4-Flash?hardware=b300&features=tool_calling,reasoning) for detailed instructions and other hardware configurations. ```bash vllm serve deepseek-ai/DeepSeek-V4-Flash-0731 \ --trust-remote-code --kv-cache-dtype fp8 --block-size 256 \ --data-parallel-size 4 --enable-expert-parallel \ --moe-backend deep_gemm_mega_moe \ --attention-config '{"use_fp4_indexer_cache": true}' \ --speculative-config '{"method":"dspark","num_speculative_tokens":7,"draft_sample_method":"greedy"}' ``` ## How to Run with SGLang Enable DSpark with `--speculative-algorithm DSPARK` and do not set a separate `--speculative-draft-model-path` as the target and draft weights therefore come from the same checkpoint. See the [SGLang cookbook](https://docs.sglang.io/cookbook/autoregressive/DeepSeek/DeepSeek-V4#hw=gb300&variant=flash-official&quant=fp4&strategy=low-latency&nodes=single) for detailed instructions, benchmarks and other hardwares configurations. ```bash sglang serve \ --trust-remote-code \ --model-path deepseek-ai/DeepSeek-V4-Flash-0731 \ --tp 4 \ --moe-runner-backend flashinfer_mxfp4 \ --speculative-algorithm DSPARK \ --mem-fraction-static 0.90 \ --chunked-prefill-size 4096 \ --swa-full-tokens-ratio 0.1 \ ``` ## How to Run Locally Please refer to the [inference](inference/README.md) folder for detailed instructions on running DeepSeek-V4 locally, including model weight conversion and interactive chat demos. For local deployment, we recommend setting the sampling parameters to `temperature = 1.0`, with `top_p = 0.95` for agentic scenarios and `top_p = 1.0` otherwise. For the `high` and `max` reasoning effort levels, we recommend a maximum output length of **384K** tokens. ## License This repository and the model weights are licensed under the [MIT License](LICENSE). ## Citation ``` @misc{deepseekai2026deepseekv4, title={DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence}, author={DeepSeek-AI}, year={2026}, } ``` ## Contact If you have any questions, please raise an issue or contact us at [service@deepseek.com](service@deepseek.com).
- Reality
- 6.5
- Singularity pressure
- 3.7/10
- Confidence
- 76%
Movement / 30D
Thirty days on the frontier
The global index recalculated daily with the same evidence, confidence and recency model.
1050
0.006 JUL
0.011 JUL
0.016 JUL
0.021 JUL
0.026 JUL
0.031 JUL
3.504 AUG
Observed change +3.5Real published proposals only Accelerating
Escalation criteria
What would make us raise the index?
The factors below explain the direction of editorial pressure. They do not alter the scoring formula.
Long-horizon autonomy
Agents completing complex, multi-stage objectives for materially longer periods with less human rescue.
Recursive leverage
AI measurably improving the code, tools, training methods or models used to build the next generation.
Compute shock
A quantified step-change in training or inference cost, throughput, memory efficiency or available scale.
Robotic generalisation
Embodied systems carrying skills into unfamiliar tasks, environments or robot bodies without bespoke retraining.
Reproducible AI Science
Novel discoveries produced through AI-led research loops and reproduced beyond the originating lab.
Method / 02 dimensions
Claims and reality travel at different speeds.
Reality Score
How much of the signal is demonstrated today.
Singularity Pressure
How strongly the signal could accelerate capability, autonomy, recursive leverage or deployment if it scales.