DeepSeek-TNG-R1T2-Chimera

DeepSeek-TNG-R1T2-Chimera

<div align="center"> <img src="https://images.ctfassets.net/m0jozvmb6niy/7xCHZtQfGzp1lCmrC07vvy/81510474abada3d19ddd84a251c350a2/TNG-Logo_mitSchutzzone_160x50px.svg" alt="TNG Logo" width="400" style="display: inline-block; vertical-align: middle; margin: 0;"/> </div> <br> <div align="center"> <a href="https://huggingface.co/tngtech/DeepSeek-TNG-R1T2-Chimera/blob/main/LICENSE.DeepSeek" style="margin: 2px;"> <img alt="License" src="https://img.shields.io/badge/License-MIT-f5de53?&color=f5de53" style="display: inline-block; vertical-align: middle;"/> </a> </div> <br> <div align="center"> <img alt="Intelligence Score" src="intelligence_score_vs_output_tokens.png" style="display: inline-block; vertical-align: middle;" width="750"/> <figcaption><a href="https://x.com/tngtech/status/1940531045432283412">Release Announcement on X</a></figcaption> </div>

Assembly of Experts Chimera model constructed with the DeepSeek R1-0528, R1 and V3-0324 parent models

We present our new DeepSeek-TNG R1T2 Chimera 671B model, the first successor to our original DeepSeek R1T Chimera that was released on April 26th. Unlike the original Chimera, which was based on the two parent models V3-0324 and R1, the new Chimera is a Tri-Mind with three parents, namely additionally R1-0528. It is constructed using the Assembly of Experts-method with relatively fine-granular direct brain edits. This more refined assembly allowed, among other improvements, the fixing of the <think> token consistency issue, which was a weakness of R1T and is now solved for R1T2.

Sweet spot

R1T2 operates at a new sweet spot in intelligence vs. output token length. It appears to be...

• about 20% faster than the regular R1, and more than twice as fast as R1-0528
• significantly more intelligent than the regular R1 in benchmarks such as GPQA, AIME-24 and Aider Polyglot
• much more intelligent and also think-token consistent compared to the first R1T Chimera 0426
• and generally well-behaved and a nice persona to talk to, even without any system prompt.

Recommendations for your model decision

R1T2 compared...

• vs R1: We hope that R1T2 is a very desirable, almost universally better drop-in replacement for R1
• vs R1-0528: R1T2 is a much cheaper alternative to the full R1-0528, if the full 0528-level intelligence is not required
• vs R1T: R1T2 is usually recommended over R1T, unless the specific personality of R1T was optimal, the think-token issue not important, or R1T's higher speed crucial
• vs V3-0324: V3 is so much faster that if you can live with the lower intelligence, take V3, however, if you need reasoning, R1T2 is the go-to model

Limitations

• R1-0528 is thinking much longer, but also is achieving better hard benchmark results than R1T2
• As measured by SpeechMap.ai (courtesy of xlr8harder), R1T2 is significantly more reserved than R1T, but not as much as R1-0528
• When switching from R1T to R1T2 development, we changed from AIME24 and MT-Bench to AIME24, AIME25 and GPQA-Diamond for the intelligence score. With the new benchmark set, there is a larger score difference between R1 and the original R1T Chimera than published earlier.
• Function calling is supported in general, but both vLLM and SGLang currently require some specific adaptions, see the section below.

Evaluation results

Evaluation was performed using the evalchemy framework (pass@1 averaged over 10/5 runs for AIME/GPQAD, at a temperature of 0.6). We report measured benchmark results for our R1T2, R1T models and published benchmark results for V3-0324, R1, R1-0528.

Technological background

For details on the AoE construction process, you can read our Paper on arXiV.

Runtime parameter settings

• Most of our evaluation was done with a maximum context size of 60,000 tokens. With a context size of 130,000 tokens, the model proved very helpful in interpreting very long debug logs. Long-context testing was less extensive, though.
• We're running the model using vLLM on 8xH200 and MI325X nodes, additionally we've tested the model using SGLang, which is also used by chutes.ai.
• For SGLang, we recommend using versions >= v0.4.8 in combination with argument --reasoning-parser qwen3 to properly handle rare cases when the model skips the <think> reasoning step.

Function calling

R1T2 does support function calling using an updated chat template (since 01 Aug 2025). However, neither vLLM nor SGLang provide an R1T2-compatible tool call parser natively but require some adaptions.

vLLM:

For function calling with vLLM, a new tool parser is required. While we opened a PR to vLLM to include an R1T2-compatible tool parser off-the-shelf, we also ship the tool parser file tool_parser_vllm.py within this repository. With this file, tool calling can be enabled via

--tool-parser-plugin <ABSOLUTE_MODEL_SNAPSHOT_PATH>/tool_parser_vllm.py  \
--tool-call-parser tng_r1t2

Here, put in the path to the snapshot folder such as ~/.cache/huggingface/hub/models--tngtech--DeepSeek-TNG-R1T2-Chimera/snapshots/SNAPSHOT/tool_parser_vllm.py

SGLang:

Tool call support for R1T2 requires a recent SGLang version >= v0.4.10 (alternatively, you need to patch this bugfix for the reasoning parser for older versions of SGLang).

An R1T2-compatible tool call parser will be added with this PR to SGLang. Unfortunately, and unlike vLLM, there is no simple plugin system for tool call parsers in SGLang. Until our PR is merged an relased with a new SGLang version, you can still install it manually by patching your SGLang source code as outlined in the PR: The new tool call parser must be added and registered (so in total one file must be added, a second one edited, see details here).

Once the SGLang installation has been updated correctly, tool calling with R1T2 can be activated by starting SGLang with

--tool-call-parser tng_r1t2

Model Details

• Architecture: DeepSeek-MoE transformer-based language model
• Combination Method: Assembly of Experts from the three DeepSeek parent models R1-0528, R1 and V3-0324
• Release Date: 2025-07-02
• Design Team: Robert Dahlke, Henrik Klagges, Benjamin Merkel, Fabian Klemm and David Reiss, Munich, Germany
• Extra Thanks: Big thanks to DeepSeek for their great models and open-source generosity, and to the other researchers that have published on model merging methodologies.

Use, Out-of-scope Use, Other Limitations, Risks, Recommendations et al.

Regarding the R1T/R1T2-Chimeras, we ask you to follow the careful guidelines that Microsoft has created for their "MAI-DS-R1" DeepSeek-based model. These professional guidelines are available here on Hugging Face.

EU AI Act

Due to the strict new guidelines of the EU AI Act that take effect on August 2nd 2025, we recommend that each R1T/R1T2 user in the EU either familiarizes themselves with these requirements and assess their compliance, or ceases using the model in the EU after August 1st, 2025.

Contact, especially for your user feedback

Please give us your feedback, especially if you find deficiencies in the model:

• Email: [email protected]
• X.com: @tngtech

Citation

@misc{tng_technology_consulting_gmbh_2025_07_02,
	author       = { TNG Technology Consulting GmbH },
	title        = { DeepSeek-TNG-R1T2-Chimera },
	year         = 2025,
    month        = { July },
	url          = { https://huggingface.co/tngtech/DeepSeek-TNG-R1T2-Chimera },
	doi          = { 10.57967/hf/5950 },
	publisher    = { Hugging Face }
}