Leaderboard / Tencent (Hunyuan)

Hunyuan 7B Instruct

Name: Hunyuan 7B Instruct
Author: Tencent (Hunyuan)

Tencent (Hunyuan) · Hunyuan · released 2025-08-04 · data checked 2026-11-01 Needs verification

Hunyuan 7B Instruct is a 7.5B-parameter dense model from Tencent (Hunyuan) (China), released 2025-08-04 under the Tencent Hunyuan Community license. It scores 52.5 on the Small LM Index, #11 of 26 small models. It needs about 6.3 GB of memory at 4-bit with an 8K-token context.

Small LM Index

52.5

#11 of 26 scored small models

Likely 48.1–56.7 · from 3 of 6 core benchmarks

Parameters

7.5B

Dense

Price per 1M tokens

—

No hosted API on OpenRouter; self-host

Context window

256K

tokens

Benchmarks

Each bar shows this model's score. The tick marks the best small model on that benchmark. Sources: 5 reported by Tencent (Hunyuan). How we label sources.

GPQA Diamond Lab-reported Knowledge60.1

Class best: Qwen3.5 9B 81.7

AIME Lab-reported Math · AIME 202575.3

Class best: Apriel 1.6 15B Thinker 88.0

LiveCodeBench Lab-reported Coding57.0

Class best: Granite 4.2 8B 73.2

BFCL Lab-reported Agentic · BFCL v370.8

MATH-500 Lab-reported Math · MATH93.7

Class best: Falcon H1R 7B 97.4

Not reported: estimated

What the index model expects from this model's other scores. Not a measurement.

MMLU-Pro≈ 74

Humanity's Last Exam≈ 7

SWE-bench Verified≈ 34

Specifications

Lab: Tencent (Hunyuan) (CN · China)
Architecture: Dense transformer
Total parameters: 7.5B
Active parameters: All (dense)
Size class: Small
Context window: 256K tokens
Input modalities: text
Reasoning: Hybrid (thinking on/off)
License: Tencent Hunyuan Community
Openness: Open weights
Memory to run: ~6.3 GB at 4-bit, 8K context (more sizes)
Hugging Face: 9.6K downloads (30 days) · 90 likes

Weights on Hugging Face ↗

Compare with

Hunyuan 7B Instruct vs Falcon H1R 7B (7.6B, index 57.5)
Hunyuan 7B Instruct vs DeepSeek-R1-Distill-Qwen-7B (7.6B, index 39.8)
Hunyuan 7B Instruct vs OLMo 3 7B Think (7.3B, index 43.4)
Hunyuan 7B Instruct vs MiMo-7B-RL (7.8B, index 40.7)

Sources

https://huggingface.co/tencent/Hunyuan-7B-Instruct

Something wrong here? Report it.

Run it locally

Uses bartowski/tencent_Hunyuan-7B-Instruct-GGUF (Q4_K_M). Swap the tag for another size: BF16, IQ2_M, IQ3_M, IQ3_XS, IQ3_XXS, IQ4_NL, IQ4_XS, Q2_K….

Ollama

ollama run hf.co/bartowski/tencent_Hunyuan-7B-Instruct-GGUF:Q4_K_M

llama.cpp

llama-server -hf bartowski/tencent_Hunyuan-7B-Instruct-GGUF:Q4_K_M

Then open http://localhost:8080 for a chat window.

MLX (Mac)

pip install -U mlx-lm
mlx_lm.chat --model mlx-community/Hunyuan-7B-Instruct-4bit

vLLM (GPU server)

vllm serve tencent/Hunyuan-7B-Instruct

Full-precision weights, for a server GPU.

Open in LM Studio

Memory needed: weights, context cache and runtime buffers.

Quantization	4K context	32K context	256K context
16-bit F16	16 GB	20 GB	49 GB
8-bit Q8_0	9.2 GB	13 GB	42 GB
4-bit Q4_K_M	5.7 GB	9.5 GB	39 GB

Check it against your GPU, Mac or phone →