Value: quality per dollar

BenchLeader Index divided by the blended price per million tokens (3 input : 1 output). Higher means more quality for the money.

As of 21 Oct 2026, BenchLeader ranks Gemma 4 E4B first on the Value: quality per dollar board at 1112.5, ahead of Granite 4.2 3B (845.3) and Gemma 4 E2B (832.0).

284 models · data as of 21 Oct 2026

Model sizeAll 99Tiny 2Small 14Medium 7Large 34

• OpenAI13
• Alibaba11
• Google9
• Mistral AI9
• Anthropic8
• DeepSeek7
• Meta6
• NVIDIA4
• Xiaomi4
• Zhipu AI4
• IBM3
• Moonshot AI3
• SpaceXAI3
• Inception2
• MiniMax2
• Upstage2
• Ant Group1
• AApodex1

The dashed staircase is the efficiency frontier: at each price, the best index anyone reaches. A model above and left of the rest gives you more quality per dollar of real work. Size bands and run costs come from Artificial Analysis, which is the only source that publishes what a fixed batch of work actually cost.

1. 1Gemma 4 E4B (thinking)1112.5
2. 2Granite 4.2 3B845.3
3. 3Gemma 4 E2B (thinking)832.0
4. 4Qwen3.5-4B (thinking)776.7
5. 5LFM2 24B A2B769.8
6. 6Llama 3.1 8B754.2
7. 7Mistral Small 3677.6
8. 8Nova Micro604.6
9. 9NVIDIA Nemotron Nano 9B v2 (thinking)592.9
10. 10DeepSeek-R1-0528-Qwen3-8B554.4
11. 11Qwen3.5 Flash554.3
12. 12Sarvam 105B (high)550.7
13. 13Gemma 3 4B549.2
14. 14Nemotron 3 Nano 30B A3B (thinking)529.5
15. 15Ling 3.0 Flash VL500.9

SearchAll providersAI21 LabsAlibabaAmazonAnt GroupAnthropicApodexArcee AIByteDanceCelerisCohereDeepSeekGoogleIBMInceptionLiquid AIMeituanMetaMicrosoftMiniMaxMistral AIMoonshot AIMultiverse ComputingNex AGINous ResearchNVIDIAOpenAIReflectionRekaSapiens AISarvamSpaceXAIStepFunSwiss AIThinking MachinesUpstageXiaomiZhipu AIOpen weightsNew models284 of 284

#	Model	Index per $ ↓	Index ↕	Blended $/M ↕
1	Gemma 4 E4BthinkingGoogleopen ↗	1112.5	44.5	$0.040
2	Granite 4.2 3BIBMopen ↗	845.3	44.8	$0.053
3	Gemma 4 E2BthinkingGoogleopen ↗	832.0	41.6	$0.050
4	Qwen3.5-4BthinkingAlibabaopen ↗	776.7	46.6	$0.060
5	LFM2 24B A2BLiquid AIopen ↗	769.8	40.8	$0.053
6	Llama 3.1 8BMetaopen ↗	754.2	36.2	$0.048
7	Mistral Small 3Mistral AIopen ↗	677.6	39.3	$0.058
8	Nova MicroAmazon	604.6	39.3	$0.065
9	NVIDIA Nemotron Nano 9B v2thinkingNVIDIAopen	592.9	41.5	$0.070
10	DeepSeek-R1-0528-Qwen3-8BDeepSeekopen ↗	554.4	37.7	$0.068
11	Qwen3.5 FlashAlibaba	554.3	52.1	$0.094
12	Sarvam 105BhighSarvamopen	550.7	40.2	$0.073
13	Gemma 3 4BGoogleopen ↗	549.2	34.6	$0.063
14	Nemotron 3 Nano 30B A3BthinkingNVIDIAopen ↗	529.5	46.6	$0.088
15	Ling 3.0 Flash VLAnt Groupopen ↗	500.9	55.6	$0.111
16	Nemotron 3.5 LightningNVIDIAopen ↗	500.0	47.5	$0.095
17	Gemma 3 12BGoogleopen ↗	493.3	37.0	$0.075
18	Llama 3-8BMetaopen	480.8	35.1	$0.073
19	Ling 3.0 Flash FinAnt Groupopen	472.2	54.3	$0.115
20	Granite 3.3 8Bno reasoningIBMopen ↗	461.2	39.2	$0.085
21	Ling 3.0 FlashAnt Groupopen ↗	459.1	52.8	$0.115
22	Llama 3.2 1B InstructMetaopen ↗	456.3	32.4	$0.071
23	Qwen2.5 7B InstructAlibabaopen ↗	443.3	39.9	$0.090
24	Granite 4.2 8BIBMopen ↗	427.8	46.2	$0.108
25	GPT-5 nanoOpenAI	405.1	55.9	$0.138
26	Agnes 2.5 Pro BetaSapiens AI	394.0	59.1	$0.150
27	gpt-oss-20bhighOpenAIopen ↗	385.2	44.3	$0.115
28	Mistral Small 3.1Mistral AIopen	382.1	36.3	$0.095
29	Mistral Small 3.2Mistral AIopen ↗	378.2	41.6	$0.110
30	Nova LiteAmazon	370.5	38.9	$0.105
31	Ministral 3 3BMistral AIopen ↗	370.0	37.0	$0.100
32	Step 3.5 FlashStepFunopen ↗	358.7	53.8	$0.150
33	MiMo-V2-FlashthinkingXiaomiopen ↗	350.0	52.5	$0.150
34	DeepSeek V4 FlashmaxDeepSeekopen ↗	349.1	61.1	$0.175
35	MiMo-V2.6-FlashXiaomiopen ↗	341.7	59.8	$0.175
36	Llama 2-7BMetaopen ↗	338.0	33.8	$0.100
37	Gemma 4 12BthinkingGoogleopen ↗	336.7	50.5	$0.150
38	GLM 4.7 FlashZhipu AIopen ↗	335.9	51.4	$0.153
39	Qwen3.5-9BthinkingAlibabaopen ↗	316.1	49.0	$0.155
40	MiMo-V2.5Xiaomiopen ↗	310.3	54.3	$0.175
41	Qwen-PlusAlibaba	307.6	48.6	$0.158
42	Claude Haiku 5.5xhighAnthropic	307.5	61.5	$0.200
43	Qwen3 8BthinkingAlibabaopen ↗	306.3	38.6	$0.126
44	Qwen3 30B A3B 2507Alibabaopen ↗	304.7	45.7	$0.150
45	GPT-6 LunamaxOpenAI	300.5	60.1	$0.200
46	Apertus 8B InstructSwiss AIopen	299.2	37.4	$0.125
47	Llama 3.2 3B InstructMetaopen ↗	294.3	35.9	$0.122
48	Gemma 4 26B A4BGoogleopen ↗	284.8	56.4	$0.198
49	Devstral Small 2507Mistral AIopen ↗	279.3	41.9	$0.150
50	Devstral Small 2505Mistral AIopen	278.0	41.7	$0.150
51	Gemini 2.5 Flash-LiteGoogle	276.0	48.3	$0.175
52	Gemini 2.5 Flash Lite 09thinkingGoogle	273.7	47.9	$0.175
53	Phi-4-MiniMicrosoftopen ↗	265.2	35.8	$0.135
54	GLM 5.3 FlashZhipu AIopen ↗	263.9	62.8	$0.238
55	Qwen3.8-Flash-NextAlibabaopen ↗	262.2	60.3	$0.230
56	Gemma 3 27BGoogleopen ↗	260.0	39.0	$0.150
57	Ministral 3 8BMistral AIopen ↗	256.0	38.4	$0.150
58	Gemma 4 31BthinkingGoogleopen ↗	253.7	52.0	$0.205
59	Ministral 8BMistral AIopen	252.7	37.9	$0.150
60	NewMiniMax-M3.1-FlashMiniMax	236.3	61.9	$0.262

Show 100 more

Cite as: BenchLeader, “Value: quality per dollar”, https://www.benchleader.com/leaderboards/value, data as of 21 Oct 2026.