GPQA Diamond

GPQA Diamond: graduate-level science multiple choice; accuracy.

404models

–Open Index weight

–last updated

Artificial Analysispublisher

#	Model	Effort	Score	Cost / task	Harness
1	GPT-6 Astra	xhigh	96.3%	–	–
2	Gemini 3.8 Flash	high	95.3%	–	–
3	Grok 4.6	high	94.9%	–	–
4	Gemini 3.7 Flash	high	94.5%	–	–
5	Muse Spark 1.3	xhigh	94.1%	–	–
6	GPT-5.6 Sol	max	94.1%	–	–
7	Gemini 3.1 Pro	–	94.1%	–	–
8	Fable 5.1	max	93.7%	–	–
9	Opus 5	xhigh	93.7%	–	–
10	Kimi K3	max	93.5%	–	–
11	Qwen 3.8 2.4T-A95B	–	93.5%	–	–
12	GPT-5.5	xhigh	93.5%	–	–
13	Grok 4.5	high	93.1%	–	–
14	MiniMax M3	–	92.9%	–	–
15	Qwen 3.8 Max	max	92.8%	–	–
16	DeepSeek V4 Pro 0813	max	92.8%	–	–
17	Gemini 3.6 Flash	high	92.8%	–	–
18	Fable 5	max	92.6%	–	–
19	GPT-5.6 Terra	max	92.5%	–	–
20	Qwen 3.8 Flash	–	92.3%	–	–
21	Qwen 3.7 Max	max	92.3%	–	–
22	Gemini 3.5 Flash	high	92.2%	–	–
23	Opus 4.8	max	92.0%	–	–
24	GPT-5.4	xhigh	92.0%	–	–
25	GLM 5.3	max	91.7%	–	–
26	GPT-5.3 Codex	xhigh	91.5%	–	–
27	Opus 4.7	max	91.4%	–	–
28	DeepSeek V4 Flash	max	91.3%	–	–
29	GLM 5.3 Flash	max	91.2%	–	–
30	Sonnet 5	max	91.1%	–	–
31	GPT-5.6 Luna	max	91.1%	–	–
32	Kimi K2.6	–	91.1%	–	–
33	Grok 4.20	–	91.1%	–	–
34	DeepSeek V4 Flash 0731	max	90.8%	–	–
35	Gemini 3 Pro	high	90.8%	–	–
36	Agnes 2.5 Pro Beta	–	90.5%	–	–
37	Qwen 3.8 27B	xhigh	90.5%	–	–
38	DeepSeek V4 Pro 0424	high	90.5%	–	–
39	Muse Spark 1.2	xhigh	90.4%	–	–
40	GPT-5.2	xhigh	90.3%	–	–
41	Grok 4.3	high	90.1%	–	–
42	Qwen 3.7 Plus	–	90.0%	–	–
43	GPT-5.2 Codex	xhigh	89.9%	–	–
44	Muse Spark 1.1	xhigh	89.8%	–	–
45	Gemini 3 Flash	–	89.8%	–	–
46	Hy3	–	89.7%	–	–
47	Opus 4.6	max	89.6%	–	–
48	Kimi K2.7 Code	–	89.6%	–	–
49	GLM 5.2	max	89.5%	–	–
50	Grok Build 0.1	–	89.5%	–	–
51	Inkling Small	–	89.5%	–	–
52	DeepSeek V4 Flash 0420	max	89.4%	–	–
53	Qwen 3.5 397B-A17B	–	89.3%	–	–
54	Nex-N2-Pro	–	89.2%	–	–
55	Solar Pro 4	–	89.1%	–	–
56	Muse Spark Open (Xhigh)	xhigh	89.0%	–	–
57	Qwen 3.6 Max	max	88.8%	–	–
58	Muse Spark	–	88.4%	–	–
59	Qwen 3.6 Plus	–	88.2%	–	–
60	Kimi K2.5	–	87.9%	–	–
61	Grok 4	–	87.7%	–	–
62	Sonnet 4.6	max	87.5%	–	–
63	GPT-5.4 Mini	xhigh	87.5%	–	–
64	MiniMax M2.7	–	87.4%	–	–
65	K2 Horizon 375B A23B	–	87.3%	–	–
66	GPT-5.1	high	87.3%	–	–
67	Inkling	xhigh	87.2%	–	–
68	DeepSeek V3.2	–	87.1%	–	–
69	MiMo V2 Pro	–	87.0%	–	–
70	Motif 3 (Beta)	–	86.9%	–	–
71	GLM 5.1	–	86.8%	–	–
72	Nemotron 3 Ultra	max	86.7%	–	–
73	Opus 4.5	–	86.6%	–	–
74	MiMo V2.5 Pro	–	86.6%	–	–
75	Apodex 1.1	–	86.4%	–	–
76	Ling 3.0 Flash	–	86.2%	–	–
77	Qwen 3 Max	max	86.1%	–	–
78	GPT-5.1 Codex	high	86.0%	–	–
79	GLM 4.7	–	85.9%	–	–
80	Qwen 3.5 27B	–	85.8%	–	–
81	Solar Open2 250B	–	85.7%	–	–
82	A.X-K2	–	85.7%	–	–
83	Qwen 3.5 122B-A10B	–	85.7%	–	–
84	Ring-2.6-1T	xhigh	85.7%	–	–
85	Gemma 4 31B	–	85.7%	–	–
86	MiMo V2	–	85.5%	–	–
87	KAT Coder Pro V2	none	85.5%	–	–
88	Ling 3.0 Flash	–	85.5%	–	–
89	GPT-5	high	85.4%	–	–
90	Grok 4.1 Fast	–	85.3%	–	–
91	MiMo V2.5	–	84.9%	–	–
92	Nanbeige4.1-3B	–	84.9%	–	–
93	MiniMax M2.5	–	84.8%	–	–
94	GLM 5	–	84.7%	–	–
95	Grok 4 Fast	–	84.7%	–	–
96	MiMo V2 Flash	–	84.6%	–	–
97	GPT-5.5 Instant	–	84.6%	–	–
98	JT-4.1 Flash 236B A21B (Non-reasoning)	none	84.5%	–	–
99	o3 pro	–	84.5%	–	–
100	Qwen 3.5 35B-A3B	–	84.5%	–	–
101	Gemini 2.5 Pro	–	84.4%	–	–
102	Qwen 3.6 27B	–	84.2%	–	–
103	Qwen 3.6 35B-A3B	–	84.1%	–	–
104	Gemini 3.5 Flash-Lite	high	83.8%	–	–
105	Kimi K2	–	83.8%	–	–
106	GPT-5 Codex	high	83.7%	–	–
107	Muse Glimmer (High)	high	83.5%	–	–
108	Motif 3	–	83.4%	–	–
109	Sonnet 4.5	–	83.4%	–	–
110	Step 3.5 Flash	–	83.1%	–	–
111	MiniMax M2.1	–	83.0%	–	–
112	K-EXAONE 2.0 0803	–	82.9%	–	–
113	JT-35B-Flash	none	82.9%	–	–
114	GPT-5 Mini	high	82.8%	–	–
115	o3	–	82.7%	–	–
116	Qwen 3.5	none	82.6%	–	–
117	Step 3.5 Flash 2603	–	82.6%	–	–
118	K2 Horizon MoVA 36B A4B	–	82.2%	–	–
119	Gemini 3.1 Flash-Lite	high	82.2%	–	–
120	GPT-5.4 Nano	xhigh	81.7%	–	–
121	GPT-5.1 Codex Mini	high	81.3%	–	–
122	DeepSeek R1	–	81.3%	–	–
123	Nova 2.0 Lite (High)	high	81.1%	–	–
124	ERNIE 4.5 300B A47B	none	81.1%	–	–
125	GLM 5 V	–	80.9%	–	–
126	Step 3.7 Flash	high	80.9%	–	–
127	Opus 4.1	–	80.9%	–	–
128	Qwen 3.5 9B	–	80.6%	–	–
129	G9v3-39A5B	–	80.5%	–	–
130	Nemotron 3 Super 120B A12B (Reasoning)	–	80.0%	–	–
131	Opus 4	–	79.6%	–	–
132	EXAONE 4.5 33B (Reasoning)	–	79.4%	–	–
133	Gemini 2.5 Flash	–	79.3%	–	–
134	Gemma 4 26B	–	79.2%	–	–
135	DeepSeek V3.1	–	79.2%	–	–
136	Grok 3 Mini	high	79.1%	–	–
137	Qwen 3 235B-A22B	–	79.0%	–	–
138	Nova 2.0 Pro Preview (Medium)	medium	78.5%	–	–
139	o4 mini	high	78.4%	–	–
140	K-EXAONE (Reasoning)	–	78.3%	–	–
141	GLM 4.5	–	78.2%	–	–
142	gpt-oss-120b	high	78.2%	–	–
143	LongCat 2.0	–	78.0%	–	–
144	GLM 4.6	–	78.0%	–	–
145	Sonnet 4	–	77.7%	–	–
146	MiniMax M2	–	77.7%	–	–
147	ERNIE 5.0 Thinking Preview	–	77.7%	–	–
148	Ring-1T	–	77.4%	–	–
149	o3 mini	high	77.3%	–	–
150	Sonnet 3.7	–	77.2%	–	–
151	Qwen 3	–	77.2%	–	–
152	Qwen 3.5 4B	–	77.1%	–	–
153	Mercury 2	high	77.0%	–	–
154	Mistral Small 4	–	76.9%	–	–
155	Nova 2.0 Lite (Medium)	medium	76.8%	–	–
156	Cogito v2.1 (Reasoning)	–	76.8%	–	–
157	o1	–	76.5%	–	–
158	KAT-Coder-Pro V1	none	76.4%	–	–
159	Doubao Seed Code	–	76.4%	–	–
160	Command A	–	76.1%	–	–
161	INTELLECT-3 (Based on GLM-4.5-Air)	–	76.1%	–	–
162	Nova 2.0 Omni (Medium)	medium	76.0%	–	–
163	Nemotron Cascade 2 30B A3B	–	75.8%	–	–
164	North Mini Code	–	75.7%	–	–
165	NVIDIA Nemotron 3 Nano 30B A3B (Reasoning)	–	75.7%	–	–
166	K2 Horizon 7B	–	75.3%	–	–
167	Gemma 4 12B	–	75.3%	–	–
168	Ling-2.6-1T	none	75.2%	–	–
169	Trinity Large Thinking	–	75.2%	–	–
170	Nova 2.0 Pro Preview (Low)	low	75.1%	–	–
171	Mistral Medium 3.5	high	74.8%	–	–
172	Llama Nemotron Super 49B v1.5 (Reasoning)	–	74.8%	–	–
173	Nemotron 3.5 Lightning	–	74.3%	–	–
174	Magistral Medium 1.2	medium	73.9%	–	–
175	EXAONE 4.0 32B (Reasoning)	–	73.9%	–	–
176	Sarvam 105B (High)	high	73.8%	–	–
177	Qwen 3 Coder	none	73.7%	–	–
178	Ling 3.0 Tiny	–	73.4%	–	–
179	Apriel-v1.6-15B-Thinker	–	73.3%	–	–
180	HyperNova 60B 2605 (High, Based on gpt-oss-120b)	high	73.3%	–	–
181	GLM 4.5 Air	–	73.3%	–	–
182	Quasar 438B (Max, Based on GLM-5.2)	max	73.2%	–	–
183	Llama 3.1	max	72.8%	–	–
184	Grok Code Fast 1	–	72.7%	–	–
185	Llama 3.1 405b	–	72.7%	–	–
186	Seed-OSS-36B-Instruct	–	72.6%	–	–
187	Ring-flash-2.0	–	72.5%	–	–
188	Solar Pro 3	high	72.4%	–	–
189	Mi:dm K 2.5 Pro Preview	–	72.2%	–	–
190	GLM 4.6 V	–	71.9%	–	–
191	Ling-1T	none	71.9%	–	–
192	Apriel-v1.5-15B-Thinker	–	71.3%	–	–
193	K2 Think V2	high	71.3%	–	–
194	Gemini 2.5 Flash Lite	–	70.9%	–	–
195	Qwen 3 30B-A3B	–	70.7%	–	–
196	MiniCPM5-2B	–	70.2%	–	–
197	Mi:dm K 2.5 Pro	–	70.1%	–	–
198	Gemini 2.0 Flash	–	70.1%	–	–
199	Nova 2.0 Omni (Low)	low	69.9%	–	–
200	Llama 3.1 70b	–	69.9%	–	–
201	Nova 2.0 Lite (Low)	low	69.8%	–	–
202	MiniMax M1	–	69.7%	–	–
203	K-EXAONE (Non-reasoning)	none	69.5%	–	–
204	Motif-2-12.7B-Reasoning	–	69.5%	–	–
205	Grok 3	none	69.3%	–	–
206	K2 Horizon 3.7B	–	69.2%	–	–
207	Step3 VL 10B	–	69.0%	–	–
208	gpt-oss-20b	high	68.8%	–	–
209	Solar Pro 2 (Reasoning)	high	68.7%	–	–
210	GLM 4.5 V	–	68.4%	–	–
211	K2-V2 (High)	high	68.1%	–	–
212	Mistral Large 3	none	68.0%	–	–
213	Magistral Medium 1	medium	67.9%	–	–
214	GPT-5 Nano	high	67.6%	–	–
215	JT-MINI	none	67.6%	–	–
216	Haiku 4.5	–	67.2%	–	–
217	Llama 4 Maverick	none	67.1%	–	–
218	DiffusionGemma 26B A4B	–	66.9%	–	–
219	Qwen 3 32B	–	66.8%	–	–
220	Qwen 3 4B	–	66.7%	–	–
221	GPT-4.1	none	66.6%	–	–
222	GPT-4.1 Mini	none	66.4%	–	–
223	Magistral Small 1.2	–	66.3%	–	–
224	Falcon-H1R-7B	–	66.1%	–	–
225	Solar Open 100B (Reasoning)	high	65.7%	–	–
226	Ling-flash-2.0	none	65.7%	–	–
227	DeepSeek V3	none	65.5%	–	–
228	GPT-4o	none	65.5%	–	–
229	Granite 4.2 30B	–	64.4%	–	–
230	Llama 3.3	–	64.3%	–	–
231	Magistral Small 1	–	64.1%	–	–
232	LongCat Flash Lite	none	63.6%	–	–
233	Nova 2.0 Pro Preview (Non-reasoning)	none	63.6%	–	–
234	Sarvam 30B (High)	high	63.3%	–	–
235	Granite 4.2 8B	–	63.1%	–	–
236	Celeris-1	none	63.1%	–	–
237	EXAONE 4.0 32B (Non-reasoning)	none	62.8%	–	–
238	Sonar Reasoning	–	62.3%	–	–
239	Gemini 2.0 Pro	none	62.2%	–	–
240	HyperCLOVA X SEED Think (32B)	–	61.5%	–	–
241	Olmo 3 32B Think	–	61.0%	–	–
242	Qwen 3 14B	–	60.4%	–	–
243	o1 mini	–	60.3%	–	–
244	Nova 2.0 Lite (Non-reasoning)	none	60.3%	–	–
245	Tri-21B-Think	–	60.1%	–	–
246	Sonnet 3.5	none	59.9%	–	–
247	K2-V2 (Medium)	medium	59.8%	–	–
248	Devstral 2	none	59.4%	–	–
249	Ling 2.6 Flash	none	59.3%	–	–
250	QwQ 32B	–	59.3%	–	–
251	Olmo 3.1 32B Think	–	59.1%	–	–
252	Gemini 1.5 Pro	none	58.9%	–	–
253	Qwen 3 8B	–	58.9%	–	–
254	Mistral Medium 3.1	none	58.8%	–	–
255	Qwen 2.5 Max	none	58.7%	–	–
256	Llama 4 Scout	none	58.7%	–	–
257	GLM 4.7 Flash	–	58.1%	–	–
258	Solar Pro 2 (Preview) (Reasoning)	high	57.8%	–	–
259	Mistral Medium 3	none	57.8%	–	–
260	Sonar Pro	none	57.8%	–	–
261	Gemma 4 E4B (Reasoning)	–	57.6%	–	–
262	Phi-4	none	57.5%	–	–
263	NVIDIA Nemotron Nano 12B v2 VL (Reasoning)	–	57.2%	–	–
264	Ministral 3 14B	none	57.2%	–	–
265	NVIDIA Nemotron Nano 9B V2 (Reasoning)	–	57.0%	–	–
266	Nova Premier	none	56.9%	–	–
267	Ling-mini-2.0	none	56.2%	–	–
268	Solar Pro 2 (Non-reasoning)	none	56.1%	–	–
269	Granite 4.2 3B	–	55.9%	–	–
270	LFM2.5-2.6B	–	55.8%	–	–
271	NVIDIA Nemotron Nano 9B V2 (Non-reasoning)	none	55.7%	–	–
272	QwQ 32B-Preview	–	55.7%	–	–
273	Nova 2.0 Omni (Non-reasoning)	none	55.5%	–	–
274	Gemma 4 E4B (Non-reasoning)	none	54.9%	–	–
275	Solar Pro 2 (Preview) (Non-reasoning)	none	54.4%	–	–
276	Gemini 2.0 Flash-Lite	none	54.2%	–	–
277	K2-V2 (Low)	low	54.1%	–	–
278	Olmo 3.1 32B Instruct	none	53.9%	–	–
279	Tri-21B-think Preview	–	53.8%	–	–
280	Devstral Small 2	none	53.2%	–	–
281	Reka Flash 3	–	52.9%	–	–
282	Olmo 3 7B Think	–	51.6%	–	–
283	Exaone 4.0 1.2B (Reasoning)	–	51.5%	–	–
284	LFM2.5-8B-A1B	–	51.3%	–	–
285	NVIDIA Nemotron 3 Nano 4B	–	51.3%	–	–
286	GPT-4.1 Nano	none	51.2%	–	–
287	Grok 2	none	51.0%	–	–
288	Pixtral Large	none	50.5%	–	–
289	Mistral Small 3.2	none	50.5%	–	–
290	Nova Pro	none	49.9%	–	–
291	Llama 3.3 70b	none	49.8%	–	–
292	Devstral Medium	none	49.2%	–	–
293	Qwen 2.5 72B	none	49.1%	–	–
294	Opus 3	none	48.9%	–	–
295	Mistral Large 2	none	48.6%	–	–
296	Granite 4.1 30B	none	48.1%	–	–
297	Llama Nemotron Super 49B v1.5 (Non-reasoning)	none	48.1%	–	–
298	LFM2 24B A2B	none	47.4%	–	–
299	Sonar	none	47.1%	–	–
300	Grok Beta	none	47.1%	–	–
301	Ministral 3 8B	none	47.1%	–	–
302	Nemotron 3 Nano Omni 30B A3B Reasoning	–	46.9%	–	–
303	Qwen 2.5 32B	none	46.6%	–	–
304	Gemini 1.5 Flash	none	46.3%	–	–
305	Mistral Small 3	none	46.2%	–	–
306	Qwen 3.5 2B	–	45.6%	–	–
307	Mistral Small 3.1	none	45.4%	–	–
308	NVIDIA Nemotron Nano 12B v2 VL (Non-reasoning)	none	43.9%	–	–
309	G9v3-3B	–	43.8%	–	–
310	Devstral Small	none	43.4%	–	–
311	Gemma 4 E2B (Reasoning)	–	43.3%	–	–
312	Nova Lite	none	43.3%	–	–
313	Granite 4.1 8B	none	43.3%	–	–
314	Llama 3.2 90b	none	43.2%	–	–
315	Gemma 3 27B	none	42.8%	–	–
316	Jamba 1.5 Large	none	42.7%	–	–
317	GPT-4o mini	none	42.6%	–	–
318	Molmo2-8B	none	42.5%	–	–
319	Mistral	none	42.4%	–	–
320	Exaone 4.0 1.2B (Non-reasoning)	none	42.4%	–	–
321	DeepSeek V2.5	none	42.3%	–	–
322	Qwen 2.5 Coder	none	41.7%	–	–
323	Granite 4.0 H Small	none	41.6%	–	–
324	Sarvam M (Reasoning, Based on Mistral Small 3.1)	high	41.6%	–	–
325	Kimi Linear 48B A3B Instruct	none	41.2%	–	–
326	Qwen 2.5 Turbo	none	41.0%	–	–
327	Haiku 3.5	none	40.8%	–	–
328	Gemma 4 E2B (Non-reasoning)	none	40.5%	–	–
329	Sonnet 3	none	40.0%	–	–
330	Olmo 3 7B Instruct	none	40.0%	–	–
331	NVIDIA Nemotron 3 Nano 30B A3B (Non-reasoning)	none	39.9%	–	–
332	Jamba 1.7 Large	none	39.0%	–	–
333	Jamba 1.6 Large	none	38.7%	–	–
334	Mistral Small	none	38.1%	–	–
335	Llama 3 70b	none	37.9%	–	–
336	Haiku 3	none	37.4%	–	–
337	Qwen 2 72B	none	37.1%	–	–
338	Nova Micro	none	35.8%	–	–
339	Ministral 3 3B	none	35.8%	–	–
340	Qwen 3 1.7B	–	35.6%	–	–
341	Mistral Large	none	35.1%	–	–
342	GPT-4	none	34.9%	–	–
343	Mistral Medium	none	34.9%	–	–
344	Gemma 3 12B	none	34.9%	–	–
345	Claude 2.0	none	34.4%	–	–
346	LFM2 8B A1B	none	34.4%	–	–
347	LFM2.5-1.2B-Thinking	–	33.9%	–	–
348	Granite 3.3 8B (Non-reasoning)	none	33.8%	–	–
349	Granite 4.0 Micro	none	33.6%	–	–
350	Jamba Reasoning 3B	–	33.3%	–	–
351	Mixtral 8x22B Instruct	none	33.2%	–	–
352	Phi-4 Mini Instruct	none	33.1%	–	–
353	DBRX Instruct	none	33.1%	–	–
354	Claude Instant	none	33.0%	–	–
355	OLMo 2 32B	none	32.8%	–	–
356	LFM 40B	none	32.7%	–	–
357	Llama 2 70b	none	32.7%	–	–
358	LFM2.5-1.2B-Instruct	none	32.6%	–	–
359	Command-R+ (Apr '24)	none	32.3%	–	–
360	Jamba 1.7 Mini	none	32.2%	–	–
361	Llama 2 13b	none	32.1%	–	–
362	Phi-3 Mini Instruct 3.8B	none	31.9%	–	–
363	Claude 2.1	none	31.9%	–	–
364	DeepSeek Coder V2 Lite Instruct	none	31.9%	–	–
365	Phi-4 Multimodal Instruct	none	31.5%	–	–
366	Granite 4.1 3B	none	31.4%	–	–
367	LFM2 2.6B	none	30.6%	–	–
368	MiniCPM-V 4.6 1.3B	none	30.5%	–	–
369	Tiny Aya Global	none	30.5%	–	–
370	Jamba 1.5 Mini	none	30.2%	–	–
371	Jamba 1.6 Mini	none	30.0%	–	–
372	GPT-3.5	none	29.7%	–	–
373	Llama 3 8b	none	29.6%	–	–
374	Gemma 3n E4B Instruct	none	29.6%	–	–
375	K2 Horizon 0.9B	–	29.3%	–	–
376	Mixtral 8x7B Instruct	none	29.2%	–	–
377	Gemma 3 4B	none	29.1%	–	–
378	Qwen 1.5 110B	none	28.9%	–	–
379	LFM2.5-VL-1.6B	none	28.9%	–	–
380	OLMo 2 7B	none	28.8%	–	–
381	Command-R (Mar '24)	none	28.4%	–	–
382	Granite 4.0 1B	none	28.1%	–	–
383	MiniCPM5-1B (Reasoning)	–	27.8%	–	–
384	Gemma 3n E4B Instruct Preview (May '25)	none	27.8%	–	–
385	Gemini 1.0 Pro	none	27.7%	–	–
386	Apertus 70B Instruct	none	27.2%	–	–
387	Llama 3.1 8b	none	27.0%	–	–
388	MiniCPM5-1B (Non-reasoning)	none	26.9%	–	–
389	Granite 4.0 H 1B	none	26.3%	–	–
390	Granite 4.0 350M	none	26.1%	–	–
391	Granite 4.0 H 350M	none	25.7%	–	–
392	Apertus 8B Instruct	none	25.6%	–	–
393	Llama 3.2 3b	none	25.5%	–	–
394	Molmo 7B-D	none	24.0%	–	–
395	Qwen 3 0.6B	–	23.9%	–	–
396	Gemma 3 1B	none	23.7%	–	–
397	Qwen 3.5 0.8B	none	23.6%	–	–
398	OpenChat 3.5 (1210)	none	23.0%	–	–
399	Gemma 3n E2B Instruct	none	22.9%	–	–
400	LFM2 1.2B	none	22.8%	–	–
401	Llama 2 7b	none	22.7%	–	–
402	Gemma 3 270M	none	22.4%	–	–
403	Llama 3.2 11b	none	22.1%	–	–
404	Llama 3.2 1b	none	19.6%	–	–