endless-frontier/BigBang-v1 · Hugging Face

Source: https://huggingface.co/endless-frontier/BigBang-v1

https://huggingface.co/endless-frontier/BigBang-v1#introductionIntroduction

As Large Language Models (LLMs) approach human expert performance, their continued development is increasingly constrained by training tasks conceived within the limits of human knowledge. We argue that open-ended capability growth requires verifiable frontier tasks: problems at the boundary of current knowledge whose solutions can be objectively evaluated through formal methods, computation, simulation, or domain-specific tools. To this end, we introduce BigBang, a general-purpose LLM evolved from Qwen 3.6 35B-A3B through efficient post-training with an adversarial, self-evolving synthetic data framework. The framework contains two core components: (1) generator agents that continually propose and solve increasingly challenging scientific and technical problems, and (2) critic agents that evaluate correctness, difficulty, scalability, and diversity, while using held-out real research tasks to calibrate the evolving synthetic-data distribution. Through iterative generator–critic interaction, the framework constructs approximately 10,000 high-difficulty post-training examples across multiple domains. Despite the modest data scale, BigBang substantially outperforms its base model across scientific research, reasoning, coding, and tool-use benchmarks, achieving aggregate performance between DeepSeek V4 Flash (284B) and DeepSeek V4 Pro (1.6T). These results demonstrate that self-evolving synthesis of verifiable frontier tasks provides a promising path toward scalable and open-ended intelligence.

https://huggingface.co/endless-frontier/BigBang-v1#main-resultsMain Results

BigBang-V1 on eight representative benchmarks spanning long-horizon search, software engineering, scientific research, and AI research. BigBang-V1 obtains the highest reported score among the selected 35B models on all eight benchmarks. It even exceeds DeepSeek V4 Pro Preview (1.6T) on FrontierScience Research, Humanity’s Last Exam, PaperBench(Code-Dev) and BioMysteryBench-HD.

BigBang-V1 on eight representative benchmarks. Click the figure to view the PDF.

https://huggingface.co/endless-frontier/BigBang-v1#benchmark-resultsBenchmark Results

Comparison of BigBang-V1 with representative closed- and open-source frontier models, together with models at the 35B scale, across benchmarks for long-horizon search, coding, scientific research, and AI research. The “\-” indicates the score is not publicly available or not tested.

BenchmarkClaude Opus 4.8Gemini 3.1 ProGPT 5.5GLM 5.2DeepSeek V4 Flash PreviewDeepSeek V4 Pro PreviewStep-3.7 FlashQwen3.6 35B-A3BNex-N2 miniAgents A1Apodex 1.0-miniBigBang V1Long-horizon SearchBrowseComp84.385.984.468.773.283.475.867.974.148.573.976.5XBench61.4-72.465.862.264.850.832.657.252.461.858.4****Coding TasksSWE-Bench Pro69.254.258.662.152.655.456.343.650.242.338.754.2SciCode-V-Sub92.3-95.184.383.790.2-56.539.064.1-68.6SciCode-V-Main78.1-90.670.368.678.1-26.615.650.0-50.0****Scientific ResearchFS-R45.224.858.352.437.740.737.211.936.838.429.646.2HLE57.951.452.254.745.148.247.236.238.446.345.350.3BioMystery-HS88.5-76.775.368.064.457.544.842.948.950.257.5BioMystery-HD42.4-23.521.623.513.711.82.05.92.05.915.7****AI ResearchMLE-Bench(Lite)63.6-59.172.740.959.140.931.818.227.327.359.1PaperBench(Code-Dev)--64.263.640.450.436.730.714.817.320.553.6

https://huggingface.co/endless-frontier/BigBang-v1#quickstartQuickstart

For streamlined integration, we recommend using BigBang-V1 via APIs. Below is a guide to use BigBang-V1 via OpenAI-compatible API.

https://huggingface.co/endless-frontier/BigBang-v1#serving-bigbang-v1Serving BigBang-V1

BigBang-V1 can be served via APIs with popular inference frameworks. In the following, we show example commands to launch OpenAI-Compatible API servers for BigBang-V1 models.

Inference efficiency and throughput vary significantly across frameworks. We recommend using the latest framework versions to ensure optimal performance and compatibility. For production workloads or high-throughput scenarios, dedicated serving engines such as SGLang, KTransformers or vLLM are strongly recommended.

The model has a default context length of 262,144 tokens. If you encounter out-of-memory (OOM) errors, consider reducing the context window. However, because BigBang-V1 leverages extended context for complex tasks, we advise maintaining a context length of at least 128K tokens to preserve thinking capabilities.

https://huggingface.co/endless-frontier/BigBang-v1#sglangSGLang

SGLangis a fast serving framework for large language models and vision language models.sglang\>=0\.5\.10is recommended for BigBang-V1, which can be installed using the following command in a fresh environment:

uv pip install sglang[all]

Seeits documentationfor more details.

The following will create API endpoints athttp://localhost:8000/v1:

• Standard Version: The following command can be used to create an API endpoint with maximum context length 262,144 tokens using tensor parallel on 8 GPUs. python -m sglang.launch_server --model-path endless-frontier/BigBang-v1 --port 8000 --tp-size 8 --mem-fraction-static 0.8 --context-length 262144 --reasoning-parser qwen3
• Tool Use: To support tool use, you can use the following command. python -m sglang.launch_server --model-path endless-frontier/BigBang-v1 --port 8000 --tp-size 8 --mem-fraction-static 0.8 --context-length 262144 --reasoning-parser qwen3 --tool-call-parser qwen3_coder
• Multi-Token Prediction (MTP): The following command is recommended for MTP: python -m sglang.launch_server --model-path endless-frontier/BigBang-v1 --port 8000 --tp-size 8 --mem-fraction-static 0.8 --context-length 262144 --reasoning-parser qwen3 --speculative-algo NEXTN --speculative-num-steps 3 --speculative-eagle-topk 1 --speculative-num-draft-tokens 4

For detailed deployment guide, see theSGLang Qwen3.5 Cookbook.

https://huggingface.co/endless-frontier/BigBang-v1#vllmvLLM

vLLMis a high-throughput and memory-efficient inference and serving engine for LLMs.vllm\>=0\.19\.0is recommended for BigBang-V1, which can be installed using the following command in a fresh environment:

uv pip install vllm --torch-backend=auto

Seeits documentationfor more details.

The following will create API endpoints athttp://localhost:8000/v1:

• Standard Version: The following command can be used to create an API endpoint with maximum context length 262,144 tokens using tensor parallel on 8 GPUs. vllm serve endless-frontier/BigBang-v1 --port 8000 --tensor-parallel-size 8 --max-model-len 262144 --reasoning-parser qwen3
• Tool Call: To support tool use, you can use the following command. vllm serve endless-frontier/BigBang-v1 --port 8000 --tensor-parallel-size 8 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder
• Multi-Token Prediction (MTP): The following command is recommended for MTP: vllm serve endless-frontier/BigBang-v1 --port 8000 --tensor-parallel-size 8 --max-model-len 262144 --reasoning-parser qwen3 --speculative-config '{"method":"qwen3_next_mtp","num_speculative_tokens":2}'
• Text-Only: The following command skips the vision encoder and multimodal profiling to free up memory for additional KV cache: vllm serve endless-frontier/BigBang-v1 --port 8000 --tensor-parallel-size 8 --max-model-len 262144 --reasoning-parser qwen3 --language-model-only

For detailed deployment guide, see thevLLM Qwen3.5 Recipe.

https://huggingface.co/endless-frontier/BigBang-v1#ktransformersKTransformers

KTransformersis a flexible framework for experiencing cutting-edge LLM inference optimizations with CPU-GPU heterogeneous computing. For running BigBang-V1 with KTransformers, see theKTransformers Deployment Guide.

https://huggingface.co/endless-frontier/BigBang-v1#hugging-face-transformersHugging Face Transformers

Hugging Face Transformers contains alightweightserver which can be used for quick testing and moderate load deployment. The latesttransformersis required for BigBang-V1:

pip install "transformers[serving]"

Seeits documentationfor more details. Please also make sure torchvision and pillow are installed.

Then, runtransformers serveto launch a server with API endpoints athttp://localhost:8000/v1; it will place the model on accelerators if available:

transformers serve endless-frontier/BigBang-v1 --port 8000 --continuous-batching