Model

Baichuan2-7B

OPEN

Baichuan Intelligence · Baichuan 2 family · released Sep 6, 2023

Early strong bilingual (zh/en) open LLM trained on 2.6T tokens; released as Base+Chat with a full technical report and 4bit quantized variant.

ReasoningCodingVisionFunction callingTool useAgentic

-15.1

Elo · rank #407

Parameters

Active params

7B (dense)

Context

4K tokens

Architecture

Dense decoder-only Transformer with RoPE, 32 layers

License

Apache-2.0 (code) + Baichuan2 Community License for weights (free commercial use if DAU<1M, non-cloud-provider, no sublicensing)

Languages

API price (in/out)

No hosted API

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

AGIEvalReasoning42.7%#28

best: OLMo 3-Think 32B · 88.2%

BIG-Bench HardReasoning41.6%#112

best: ERNIE 4.5 300B-A47B · 94.3%

C-EvalKnowledge54.0%#49

best: Qwen3.6-Plus · 93.3%

CMMLUKnowledge57.1%#26

best: Doubao-1.5-Pro · 90.9%

GSM8KMath24.5%#162

best: Llama 3.1 405B · 96.8%

HumanEvalCoding18.3%#170

best: Claude Opus 4.5 · 99.4%

MATH-500Math5.6%#196

best: GPT-5 · 99.4%

MBPPCoding24.2%#102

best: Llama-3.3-Nemotron-Super-49B v1 (Reasoning On) · 91.3%

MMLUKnowledge54.2%#225

best: OpenAI o3 · 92.9%

Run it locally

VRAM @ Q4

4.5 GB

VRAM @ FP16

14 GB

Fits on (Q4)

RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

Throughput data unavailable.

Quantizations

GGUF · GPTQ · INT4 (bitsandbytes NF4)

Baichuan 2 family

Elo progression across releases

Baichuan2-13BSep 2023121.1 Baichuan2-7BSep 2023-15.1

Resources

HFbaichuan-inc/Baichuan2-7B-Chat GHbaichuan-inc/Baichuan2

API price weights · each benchmark row carries its own source badge (see methodology)