Benchmark suiteLatest source data: Jul 16, 2026Checked: July 20, 2026

Model value rankings.

A focused view of which frontier models deliver the strongest benchmark signal for the money, balancing quality, blended token price, latency, and output speed.

Ranked models

92

Require quality and price matches

Best value

Muse Spark 1.1

79 value index

Lowest price

$0.063

Granite 4.1 8b

Fastest TTFT

0.43s

Granite 4.1 8b

Value formula

Quality first, then cost and speed.

The value index normalizes each available signal before blending benchmark quality, blended price, time to first token, and output speed. It avoids raw score-per-dollar math, which can wildly overstate very cheap models.

Best models for the money

Higher value index is better. Missing source rows are omitted from the value ranking rather than treated as zero.

Top 30 of 92

#1
Muse Spark 1.1

muse-spark-1.1

MetaProprietary
79value index
Price
$2.00blended / 1M tokens
Latency
0.83stime to first token
Speed
113 tok/soutput speed
Quality
81quality signal
ArenaArtificial Analysis
#2
Kimi K3

kimi-k3

Moonshot AIProprietary
78value index
Price
$6.00blended / 1M tokens
Latency
1.36stime to first token
Speed
63 tok/soutput speed
Quality
87quality signal
ArenaArtificial Analysis
#3
Claude Opus 4.7

claude-opus-4-7

AnthropicProprietary
74value index
Price
$10.00blended / 1M tokens
Latency
1.61stime to first token
Speed
45 tok/soutput speed
Quality
87quality signal
ArenaArtificial Analysis
#4
Gemini 3.5 Flash High

gemini-3.5-flash-high

GoogleProprietary
74value index
Price
$3.38blended / 1M tokens
Latency
0.77stime to first token
Speed
232 tok/soutput speed
Quality
71quality signal
ArenaArtificial Analysis
#5
Claude Opus 4.6 Thinking

claude-opus-4-6-thinking

AnthropicProprietary
74value index
Price
$10.00blended / 1M tokens
Latency
11.5stime to first token
Speed
46 tok/soutput speed
Quality
88quality signal
ArenaArtificial Analysis
#6
Gemini 3 Pro

gemini-3-pro

GoogleProprietary
74value index
Price
$4.50blended / 1M tokens
Latency
Not listedtime to first token
Speed
Not listedoutput speed
Quality
67quality signal
ArenaArtificial Analysis
#7
Claude Opus 4.6

claude-opus-4-6

AnthropicProprietary
73value index
Price
$10.00blended / 1M tokens
Latency
1.82stime to first token
Speed
43 tok/soutput speed
Quality
85quality signal
ArenaArtificial Analysis
#8
GLM 5.2

glm-5.2 (max)

Z.aiOpen weights
71value index
Price
$3.16blended / 1M tokens
Latency
0.95stime to first token
Speed
81 tok/soutput speed
Quality
69quality signal
ArenaArtificial Analysis
#9
Gemini 3.5 Flash Medium

gemini-3.5-flash-medium

GoogleProprietary
71value index
Price
$3.38blended / 1M tokens
Latency
10.6stime to first token
Speed
268 tok/soutput speed
Quality
67quality signal
ArenaArtificial Analysis
#10
GLM 5.1

glm-5.1

Z.aiOpen weights
71value index
Price
$2.15blended / 1M tokens
Latency
1.01stime to first token
Speed
62 tok/soutput speed
Quality
66quality signal
ArenaArtificial Analysis
#11
GPT-5.6 Sol xHigh

gpt-5.6-sol-xhigh

OpenAIProprietary
70value index
Price
$11.25blended / 1M tokens
Latency
2.50stime to first token
Speed
60 tok/soutput speed
Quality
82quality signal
ArenaArtificial Analysis
#12
Kimi K2.7 Code

kimi-k2.7-code

Moonshot AIOpen weights
70value index
Price
$1.71blended / 1M tokens
Latency
1.16stime to first token
Speed
46 tok/soutput speed
Quality
64quality signal
ArenaArtificial Analysis
#13
Qwen3.7 Max Preview

qwen3.7-max-preview

AlibabaProprietary
69value index
Price
$3.75blended / 1M tokens
Latency
1.54stime to first token
Speed
202 tok/soutput speed
Quality
64quality signal
ArenaArtificial Analysis
#14
Claude Sonnet 5 High

claude-sonnet-5-high

AnthropicProprietary
69value index
Price
$4.00blended / 1M tokens
Latency
1.52stime to first token
Speed
62 tok/soutput speed
Quality
66quality signal
ArenaArtificial Analysis
#15
Grok 4.5

grok-4.5

xAIProprietary
68value index
Price
$3.00blended / 1M tokens
Latency
13.4stime to first token
Speed
73 tok/soutput speed
Quality
65quality signal
ArenaArtificial Analysis
#16
MiMo V2.5 Pro

mimo-v2.5-pro

XiaomiOpen weights
68value index
Price
$0.544blended / 1M tokens
Latency
2.21stime to first token
Speed
63 tok/soutput speed
Quality
58quality signal
ArenaArtificial Analysis
#17
Gemini 3.1 Pro Preview

gemini-3.1-pro-preview

GoogleProprietary
68value index
Price
$4.50blended / 1M tokens
Latency
23.7stime to first token
Speed
126 tok/soutput speed
Quality
67quality signal
ArenaArtificial Analysis
#18
Kimi K2.6

kimi-k2.6

Moonshot AIOpen weights
67value index
Price
$1.71blended / 1M tokens
Latency
1.08stime to first token
Speed
52 tok/soutput speed
Quality
59quality signal
ArenaArtificial Analysis
#19
Claude Fable 5

claude-fable-5

AnthropicProprietary
66value index
Price
$20.00blended / 1M tokens
Latency
55.9stime to first token
Speed
58 tok/soutput speed
Quality
98quality signal
ArenaArtificial Analysis
#20
Claude Opus 4.8

claude-opus-4-8

AnthropicProprietary
66value index
Price
$10.00blended / 1M tokens
Latency
21.8stime to first token
Speed
55 tok/soutput speed
Quality
75quality signal
ArenaArtificial Analysis
#21
Claude Sonnet 4.6

claude-sonnet-4-6

AnthropicProprietary
66value index
Price
$6.00blended / 1M tokens
Latency
47.9stime to first token
Speed
58 tok/soutput speed
Quality
71quality signal
ArenaArtificial Analysis
#22
Qwen3.7 Plus

qwen3.7-plus

AlibabaProprietary
66value index
Price
$0.700blended / 1M tokens
Latency
1.62stime to first token
Speed
53 tok/soutput speed
Quality
54quality signal
ArenaArtificial Analysis
#23
GPT-5.5 High

gpt-5.5-high

OpenAIProprietary
65value index
Price
$11.25blended / 1M tokens
Latency
12.5stime to first token
Speed
69 tok/soutput speed
Quality
74quality signal
ArenaArtificial Analysis
#24
Gemini 3 Flash

gemini-3-flash (thinking-minimal)

GoogleProprietary
65value index
Price
$1.13blended / 1M tokens
Latency
0.70stime to first token
Speed
212 tok/soutput speed
Quality
51quality signal
ArenaArtificial Analysis
#25
DeepSeek V4 Pro

deepseek-v4-pro

DeepSeekOpen weights
65value index
Price
$0.544blended / 1M tokens
Latency
0.95stime to first token
Speed
63 tok/soutput speed
Quality
52quality signal
ArenaArtificial Analysis
#26
DeepSeek V4 Pro Thinking

deepseek-v4-pro-thinking

DeepSeekOpen weights
65value index
Price
$0.544blended / 1M tokens
Latency
0.95stime to first token
Speed
60 tok/soutput speed
Quality
52quality signal
ArenaArtificial Analysis
#27
Minimax M2.7

minimax-m2.7

MiniMaxOpen weights
65value index
Price
$0.525blended / 1M tokens
Latency
1.16stime to first token
Speed
55 tok/soutput speed
Quality
52quality signal
ArenaArtificial Analysis
#28
Minimax M2.1 Preview

minimax-m2.1-preview

MiniMaxOpen weights
64value index
Price
$0.525blended / 1M tokens
Latency
1.14stime to first token
Speed
74 tok/soutput speed
Quality
51quality signal
ArenaArtificial Analysis
#29
Qwen3.6 Max Preview

qwen3.6-max-preview

AlibabaProprietary
64value index
Price
$2.93blended / 1M tokens
Latency
2.15stime to first token
Speed
39 tok/soutput speed
Quality
56quality signal
ArenaArtificial Analysis
#30
Qwen3.5 122b A10B

qwen3.5-122b-a10b

AlibabaOpen weights
64value index
Price
$1.10blended / 1M tokens
Latency
1.08stime to first token
Speed
149 tok/soutput speed
Quality
50quality signal
ArenaArtificial Analysis

Benchmark guide

What the scores mean.

A quick reading key for interpreting the value rankings without over-weighting one benchmark, price point, or speed measurement.

Higher: value indexLower: price + latency
What does the value index measure?

The value index blends normalized benchmark quality, blended token price, latency, and output speed. It is designed to answer which models deliver strong public performance for the money, not which model is best in every task.

Why not use raw score divided by price?

Raw score-per-dollar can exaggerate extremely cheap models and make small benchmark differences look larger than they are. Normalizing each signal keeps the ranking easier to compare across quality, price, and speed.

Why are some models missing from value rankings?

A model needs both a benchmark quality signal and a confident price match to be ranked. If the price or benchmark source is missing, the model is omitted rather than treated as free or zero quality.

How should I use price, latency, and speed together?

Price estimates cost, latency measures time to first token, and speed measures output throughput after generation begins. The best model for a product depends on which constraint matters most in the actual workflow.