olmo-3 - Ollama 框架

Olmo 是一系列开放语言模型，旨在推动语言模型科学的发展。这些模型在 Dolma 3 数据集上进行预训练，并在 Dolci 数据集上进行后训练。

Olmo 3，一个全新的 7B 和 32B 模型系列，包含 Instruct 和 Think 变体。它具有长链思维能力，以提高数学和编码等推理任务的性能。

Olmo 是一系列开放语言模型，旨在推动语言模型科学的发展。这些模型在 Dolma 3 数据集上进行预训练，并在 Dolci 数据集上进行后训练。Allen AI 团队发布了所有代码、检查点、日志和相关的训练细节。

Olmo 3 Instruct 7B

ollama run olmo-3:7b-instruct

Olmo 3 Think 7B

ollama run olmo-3:7b-think

Olmo 3 Think 32B

ollama run olmo-3:32b-think

Olmo 3 Instruct 7B

基准测试	Olmo3 Instruct 7B	Qwen 3 8B (无推理)	Qwen 3 VL 8B Instruct	Qwen 2.5 7B	Olmo 2 7B Instruct	Apertus 8B Instruct	Granite 3.3 8B Instruct
MATH	87.3	82.3	91.6	71	30.1	21.9	67.3
AIME 2024	44.3	26.2	55.1	11.3	1.3	0.5	7.3
AIME 2025	32.5	21.7	43.3	6.3	0.4	0.2	6.3
OMEGA	28.9	20.5	32.3	13.7	5.2	5.0	10.7
BigBenchHard	71.2	73.7	85.6	68.8	43.8	42.2	61.2
ZebraLogic	32.9	25.4	64.3	10.7	5.3	5.3	17.6
AGI Eval English	64.4	76	84.5	69.8	56.1	50.8	64.0
HumanEvalPlus	77.2	79.8	82.9	74.9	25.8	34.4	64.0
MBPP+	60.2	64.4	66.3	62.6	40.7	42.1	54.0
LiveCodeBench v3	29.5	53.2	55.9	34.5	7.2	7.8	11.5
IFEval	85.6	86.3	87.8	73.4	72.2	71.4	77.5
IFBench	32.3	29.3	34	28.4	26.7	22.1	22.3
MMLU	69.1	80.4	83.6	77.2	61.6	62.7	63.5
PopQA	14.1	20.4	26.5	21.5	25.5	25.5	28.9
GPQA	40.4	44.6	51.1	35.6	31.3	28.8	33.0
AlpacaEval 2 LC	40.9	49.8	73.5	23	18.3	8.1	28.6
SimpleQA	79.3	79	90.3	78	–	–	–
LitQA2	38.2	39.6	30.7	29.8	–	–	–
BFCL	49.8	60.2	66.2	55.8	–	–	–
安全	87.3	78	80.2	73.4	93.1	72.2	73.7

Olmo 3 Think 7B

基准测试	Olmo 3 Think 7B	OpenThinker3-7B	Nemotron-Nano-9B-v2	DeepSeek-R1-Distill-Qwen-7B	Qwen 3 8B (推理)	Qwen 3 VL 8B Thinker	OpenReasoning Nemotron 7B
MATH	95.1	94.5	94.4	87.9	95.1	95.2	94.6
AIME 2024	71.6	67.7	72.1	54.9	74.0	70.9	77.0
AIME 2025	64.6	57.2	58.9	40.2	67.8	61.5	73.1
OMEGA	37.8	38.4	42.4	28.5	43.4	38.1	43.2
BBH	86.6	77.1	86.2	73.5	84.4	86.8	81.3
ZebraLogic	66.5	34.9	60.8	26.1	85.2	91.2	22.4
AGI Eval	81.5	78.6	83.1	69.5	87.0	90.1	81.4
HumanEval+	89.9	87.4	89.7	83.0	80.2	83.7	89.7
MBPP+	64.7	61.4	66.1	63.5	69.1	63.0	61.2
LCB v3	75.2	68.0	83.4	58.8	86.2	85.5	82.3
IFEval	88.2	51.7	86.0	59.6	87.4	85.5	42.5
IFBench	41.6	23.0	34.6	16.7	37.1	40.4	23.4
MMLU	77.8	77.4	84.3	67.9	85.4	86.5	80.7
PopQA	23.7	18.0	17.9	12.8	24.3	29.3	14.5
GPQA	46.2	47.6	56.2	54.4	57.7	61.5	56.6
AE 2	52.1	24.0	58.0	7.7	60.5	73.5	8.6
	70.7	31.3	72.1	54.0	68.3	82.9	30.3

Olmo 3 Think 32B

基准测试	Olmo 3 Think 32B	Qwen 3 32B	Qwen 3 VL 32B Thinking	Qwen 2.5 32B	Gemma 3 27B Instruct	Gemma 2 27B Instruct	Olmo 2 32B Instruct	DeepSeek-R1-Distill-Qwen-32B
数学
MATH	96.1	95.4	96.7	80.2	87.4	51.5	49.2	92.6
AIME 2024	76.8	80.8	86.3	15.7	28.9	4.7	4.6	70.3
AIME 2025	72.5	70.9	78.8	13.4	22.9	0.9	0.9	56.3
OMEGA	50.8	47.7	50.8	19.2	24.0	9.1	9.8	38.9
推理
BigBenchHard	89.8	90.6	91.1	80.9	82.4	66.0	65.6	89.7
ZebraLogic	76.0	88.3	96.1	24.1	24.8	17.2	13.3	69.4
AGI Eval English	88.2	90.0	92.2	78.9	76.9	70.9	68.4	88.1
编码
HumanEvalPlus	91.4	91.2	90.6	82.6	79.2	67.5	44.4	92.3
MBPP+	68.0	70.6	66.2	66.6	65.7	61.2	49.0	70.1
LiveCodeBench v3	83.5	90.2	84.8	49.9	39.0	28.7	10.6	79.5
IF
IFEval	89.0	86.5	85.5	81.9	85.4	62.1	85.8	78.7
IFBench	47.6	37.3	55.1	36.7	31.3	27.8	36.4	23.8
知识 & 问答
MMLU	85.4	88.8	90.1	84.6	74.6	76.1	77.1	88.0
PopQA	31.9	30.7	32.2	28.0	30.2	30.4	37.2	26.7
GPQA	58.1	67.3	67.4	44.6	45.0	39.9	36.4	61.8
聊天
AlpacaEval 2 LC	74.2	75.6	80.9	81.9	65.5	39.8	38.0	26.2
安全	68.8	69.0	82.7	81.9	68.6	74.3	83.8	63.6

Olmo 是一系列开放语言模型，旨在推动语言模型科学的发展。这些模型在 Dolma 3 数据集上进行预训练，并在 Dolci 数据集上进行后训练。