AI Models

64 foundation models reviewed by our six-persona AI panel — benchmarks, real token pricing, context windows, and the verdicts vendors won't give you.

Sort

64 of 64 models

Anthropic · Fable

Claude Fable 5

9.5
AI Panel
Value 6.5
GA2026FrontierReasoningCoding

Best for hardest long-horizon work money can buy

$10
in / Mtok
$50
out / Mtok
1M
context
tok/s
Anthropic · Opus

Claude Opus 4.8

9.2
AI Panel
Value 8.0
GA2026FrontierReasoningCoding

Best for production agentic coding workhorse

$5
in / Mtok
$25
out / Mtok
1M
context
60
tok/s
Anthropic · Opus

Claude Opus 4.7

9.0
AI Panel
Value 7.5
GA2026FrontierReasoningCoding

Best for agentic coding at the frontier

$5
in / Mtok
$25
out / Mtok
1M
context
55
tok/s
OpenAI · mini

GPT-5.4 mini

9.0
AI Panel
Value 9.8
GA2026CodingagenticCost-Optimized

Best for best price-to-capability in the OpenAI lineup

$0.75
in / Mtok
$4.5
out / Mtok
400K
context
180
tok/s
Anthropic · Sonnet

Claude Sonnet 4.6

8.8
AI Panel
Value 9.0
GA2026FrontierCodingMultimodal

Best for best value production workhorse

$3
in / Mtok
$15
out / Mtok
1M
context
75
tok/s
Z.ai · Flagship

GLM-5.2

8.8
AI Panel
Value 9.6
GA2026Open-weightsFrontierReasoningCoding

Best for open-weights frontier coding agents

$1.4
in / Mtok
$4.4
out / Mtok
1M
context
167
tok/s
DeepSeek · Flash

DeepSeek V4-Flash

8.7
AI Panel
Value 9.9
preview2026Open-weightsReasoningCodingCost-Optimized

Best for frontier-adjacent quality at the lowest cost in market

$0.14
in / Mtok
$0.28
out / Mtok
1M
context
tok/s
OpenAI · Medium

GPT-5.4

8.7
AI Panel
Value 8.5
GA2026FrontierReasoningCoding

Best for default production workhorse

$2.5
in / Mtok
$15
out / Mtok
1.1M
context
86
tok/s
OpenAI · Pro

GPT-5.5

8.7
AI Panel
Value 6.5
GA2026FrontierReasoningCoding

Best for frontier agentic coding and computer use

$5
in / Mtok
$30
out / Mtok
1.1M
context
69
tok/s
Google · Pro

Gemini 3.1 Pro

8.7
AI Panel
Value 8.5
preview2026FrontierReasoningLong-Context

Best for frontier reasoning + long-context on Google Cloud

$2
in / Mtok
$12
out / Mtok
1.0M
context
143
tok/s
Anthropic · Opus

Claude Opus 4.6

8.6
AI Panel
Value 7.8
GA2026FrontierReasoningCoding

Best for prior frontier, stable production target

$5
in / Mtok
$25
out / Mtok
1M
context
46
tok/s
DeepSeek · Pro

DeepSeek V4-Pro

8.5
AI Panel
Value 9.7
preview2026Open-weightsFrontierReasoningCoding

Best for frontier-grade agentic coding at open-weights cost

$0.435
in / Mtok
$0.87
out / Mtok
1M
context
tok/s
Google · Flash

Gemini 3.5 Flash

8.5
AI Panel
Value 8.0
GA2026CodingMultimodalLong-Context

Best for production agent + coding backbone

$1.5
in / Mtok
$9
out / Mtok
1.0M
context
204
tok/s
Mistral AI · Small

Mistral Small 4

8.5
AI Panel
Value 9.5
GA2026Open-weightsCost-OptimizedOpen-WeightsReasoning

Best for best price-to-capability open multimodal model

$0.15
in / Mtok
$0.6
out / Mtok
256K
context
180
tok/s
NVIDIA · Ultra

Nemotron 3 Ultra

8.5
AI Panel
Value 8.5
GA2026Open-weightsFrontierReasoningCoding

Best for self-hostable 1M-context reasoning

in / Mtok
out / Mtok
1M
context
tok/s
Alibaba Cloud · Large

Qwen3-235B-A22B

8.5
AI Panel
Value 9.5
GA2025Open-weightsFrontierReasoningCoding

Best for frontier-adjacent reasoning at open-weight prices

$0.2
in / Mtok
$0.6
out / Mtok
131K
context
68
tok/s
Anthropic · Haiku

Claude Haiku 4.5

8.4
AI Panel
Value 9.8
GA2025Cost-OptimizedCodingMultimodal

Best for high-volume low-latency worker

$1
in / Mtok
$5
out / Mtok
200K
context
102
tok/s
Mistral AI · Medium

Mistral Medium 3.5

8.4
AI Panel
Value 8.5
GA2026Open-weightsFrontierCodingOpen-Weights

Best for open-weight agentic coding at sub-frontier price

$1.5
in / Mtok
$7.5
out / Mtok
256K
context
tok/s
Alibaba Cloud · Coder

Qwen2.5-Coder-32B-Instruct

8.4
AI Panel
Value 9.5
GA2024Open-weightsCodingOpen-WeightsCost-Optimized

Best for canonical self-hosted code model

$0.08
in / Mtok
$0.24
out / Mtok
131K
context
tok/s
Alibaba Cloud · Large

Qwen3-32B

8.4
AI Panel
Value 9.5
GA2025Open-weightsReasoningCodingOpen-Weights

Best for best single-GPU open-weight model in production

$0.08
in / Mtok
$0.28
out / Mtok
131K
context
tok/s
Alibaba Cloud · Max

Qwen3.7-Max

8.3
AI Panel
Value 7.5
GA2026FrontierReasoningCoding

Best for 1M-context multimodal agent flagship

$2.5
in / Mtok
$7.5
out / Mtok
1M
context
195
tok/s
DeepSeek · Large

DeepSeek V3.2

8.2
AI Panel
Value 9.3
GA2025Open-weightsReasoningCost-OptimizedOpen-Weights

Best for open-weights math/reasoning at GA stability

$0.252
in / Mtok
$0.378
out / Mtok
128K
context
tok/s
OpenAI · nano

GPT-5.4 nano

8.2
AI Panel
Value 9.9
GA2026Cost-OptimizedEdgeagentic

Best for cheapest viable frontier-family worker tier

$0.2
in / Mtok
$1.25
out / Mtok
400K
context
250
tok/s
Anthropic · Opus

Claude Opus 4.5

8.0
AI Panel
Value 7.5
GA2025FrontierReasoningCoding

Best for the Opus price reset, stable target

$5
in / Mtok
$25
out / Mtok
200K
context
53
tok/s
DeepSeek · Reasoning

DeepSeek R1 (0528)

8.0
AI Panel
Value 9.2
GA2025Open-weightsReasoningOpen-WeightsCost-Optimized

Best for exposed-CoT reasoning at a fraction of o-series cost

$0.55
in / Mtok
$2.19
out / Mtok
128K
context
tok/s
Mistral AI · Large

Mistral Large 3

8.0
AI Panel
Value 9.0
GA2025Open-weightsFrontierOpen-Weightsmultilingual

Best for EU-sovereign open-weight frontier generalist

$0.5
in / Mtok
$1.5
out / Mtok
256K
context
51
tok/s
Mistral AI · Ministral

Ministral 3 14B

7.9
AI Panel
Value 9.0
GA2025Open-weightsEdgeOpen-WeightsReasoning

Best for best small open model for on-device reasoning

$0.2
in / Mtok
$0.2
out / Mtok
256K
context
tok/s
Alibaba Cloud · VL

Qwen2.5-VL-72B-Instruct

7.9
AI Panel
Value 8.5
GA2025Open-weightsMultimodalOpen-Weights

Best for best open-weight VLM for document AI

$0.7
in / Mtok
$0.7
out / Mtok
131K
context
tok/s
Google · Flash-Lite

Gemini 3.1 Flash-Lite

7.8
AI Panel
Value 9.5
preview2026Cost-OptimizedLong-ContextMultimodal

Best for best $/intelligence in Google's lineup

$0.25
in / Mtok
$1.5
out / Mtok
1.0M
context
332
tok/s
Alibaba Cloud · Medium

Qwen3-14B

7.8
AI Panel
Value 9.5
GA2025Open-weightsOpen-WeightsCost-OptimizedEdge

Best for best open-weight model in the 13-15B band

$0.06
in / Mtok
$0.2
out / Mtok
131K
context
tok/s
DeepSeek · Large

DeepSeek V3.1

7.7
AI Panel
Value 8.8
GA2025Open-weightsReasoningCodingCost-Optimized

Best for the model that put DeepSeek on the production-agent map

$0.21
in / Mtok
$0.79
out / Mtok
128K
context
tok/s
OpenAI · Pro

GPT-5.5 Pro

7.7
AI Panel
Value 5.0
GA2026FrontierReasoning

Best for frontier deep-research and hardest reasoning

$30
in / Mtok
$180
out / Mtok
1.1M
context
tok/s
Google · Flash-Lite

Gemini 2.5 Flash-Lite

7.7
AI Panel
Value 9.5
GA2025Cost-OptimizedEdgeLong-Context

Best for cheapest GA model with 1M context

$0.1
in / Mtok
$0.4
out / Mtok
1.0M
context
215
tok/s
xAI · Flagship

Grok 4.3

7.7
AI Panel
Value 8.5
GA2026FrontierReasoningMultimodal

Best for real-time research and agentic work at frontier-class value

$1.25
in / Mtok
$2.5
out / Mtok
1M
context
182
tok/s
Moonshot AI · Coder

Kimi K2.7-Code

7.7
AI Panel
Value 8.8
GA2026Open-weightsCodingReasoningMultimodal

Best for budget long-horizon coding agents

$0.95
in / Mtok
$4
out / Mtok
262K
context
tok/s
Meta · Maverick

Llama 4 Maverick

7.7
AI Panel
Value 9.0
GA2025Open-weightsOpen-WeightsMultimodalCost-Optimized

Best for self-hosted multimodal workhorse with no vendor lock-in

$0.2
in / Mtok
$0.85
out / Mtok
1M
context
104
tok/s
Mistral AI · Medium

Mistral Medium 3.1

7.7
AI Panel
Value 9.0
GA2025Cost-OptimizedMultimodalmultilingual

Best for cheap multilingual multimodal chat at volume

$0.4
in / Mtok
$2
out / Mtok
131K
context
45
tok/s
Mistral AI · Coder

Codestral (25.08)

7.6
AI Panel
Value 8.5
GA2025CodingCost-Optimized

Best for low-latency code completion and FIM

$0.3
in / Mtok
$0.9
out / Mtok
128K
context
tok/s
Google · Pro

Gemini 2.5 Pro

7.6
AI Panel
Value 7.5
GA2025Long-ContextMultimodalReasoning

Best for cost-effective long-context Pro

$1.25
in / Mtok
$10
out / Mtok
1.0M
context
144
tok/s
Anthropic · Sonnet

Claude Sonnet 4.5

7.5
AI Panel
Value 8.0
GA2025CodingMultimodalCost-Optimized

Best for stable legacy workhorse

$3
in / Mtok
$15
out / Mtok
200K
context
60
tok/s
Meta · Scout

Llama 4 Scout

7.5
AI Panel
Value 9.5
GA2025Open-weightsOpen-WeightsMultimodalCost-Optimized

Best for single-GPU long-context open-weights deploy

$0.1
in / Mtok
$0.34
out / Mtok
10M
context
106
tok/s
Mistral AI · Ministral

Ministral 3 8B

7.5
AI Panel
Value 9.0
GA2025Open-weightsEdgeOpen-WeightsMultimodal

Best for fast multilingual edge model with vision

$0.15
in / Mtok
$0.15
out / Mtok
256K
context
tok/s
Alibaba Cloud · Large

Qwen2.5-72B-Instruct

7.5
AI Panel
Value 8.5
GA2024Open-weightsOpen-WeightsCost-Optimized

Best for mature multilingual open-weight workhorse

$0.12
in / Mtok
$0.3
out / Mtok
131K
context
tok/s
Cohere · Flagship

Command A+

7.4
AI Panel
Value 7.5
GA2026Open-weightsReasoningMultimodalOpen-Weights

Best for sovereign self-hosted enterprise agents

in / Mtok
out / Mtok
131K
context
tok/s
Mistral AI · Coder

Devstral 2

7.4
AI Panel
Value 9.0
GA2025Open-weightsCodingOpen-WeightsCost-Optimized

Best for budget open-weight agentic coding

$0.4
in / Mtok
$0.9
out / Mtok
256K
context
tok/s
Meta · Small

Llama 3.1 8B

7.4
AI Panel
Value 9.5
GA2024Open-weightsOpen-WeightsCost-OptimizedEdge

Best for on-device and high-volume open-weights workhorse

$0.05
in / Mtok
$0.08
out / Mtok
128K
context
159
tok/s
Meta · Large

Llama 3.3 70B

7.4
AI Panel
Value 8.5
GA2024Open-weightsOpen-WeightsCost-Optimized

Best for operationally-mature text-only open default

$0.12
in / Mtok
$0.4
out / Mtok
128K
context
82
tok/s
Meta · Spark

Muse Spark

7.3
AI Panel
Value 6.0
GA2026FrontierReasoningMultimodal

Best for frontier reasoning for consumers, not yet for builders

in / Mtok
out / Mtok
262K
context
tok/s
DeepSeek · VL

DeepSeek-VL2

7.2
AI Panel
Value 8.8
GA2024Open-weightsMultimodalOpen-WeightsCost-Optimized

Best for open-weights OCR and document understanding at the edge

in / Mtok
out / Mtok
4K
context
tok/s
Mistral AI · Reasoning

Magistral Medium 1.2

7.2
AI Panel
Value 7.0
GA2025ReasoningMultimodal

Best for auditable multilingual reasoning

$2
in / Mtok
$5
out / Mtok
131K
context
39
tok/s
Alibaba Cloud · Large

Qwen2.5-32B-Instruct

7.2
AI Panel
Value 8.5
GA2024Open-weightsOpen-WeightsCost-Optimized

Best for mature Apache-2.0 single-GPU workhorse

$0.1
in / Mtok
$0.25
out / Mtok
131K
context
tok/s
Mistral AI · Ministral

Ministral 3 3B

7.1
AI Panel
Value 9.0
GA2025Open-weightsEdgeOpen-WeightsMultimodal

Best for on-device AI with native vision

$0.1
in / Mtok
$0.1
out / Mtok
256K
context
tok/s
xAI · Reasoning

Grok 4.20 Multi-Agent

7.0
AI Panel
Value 6.0
GA2026ReasoningLong-Context

Best for single-call parallel deep research with live X data

$1.25
in / Mtok
$2.5
out / Mtok
2M
context
tok/s
xAI · Reasoning

Grok 4.20

6.9
AI Panel
Value 7.0
GA2026ReasoningLong-Context

Best for runtime reasoning toggle with long context and live X data

$1.25
in / Mtok
$2.5
out / Mtok
1M
context
171
tok/s
Alibaba Cloud · Reasoning

QwQ-32B

6.8
AI Panel
Value 8.0
GA2025Open-weightsReasoningOpen-Weights

Best for open-weight always-on reasoning at 32B

$0.12
in / Mtok
$0.18
out / Mtok
131K
context
tok/s
OpenAI · Pro

GPT-5.4 Pro

6.7
AI Panel
Value 4.5
GA2026FrontierReasoning

Best for escalation tier for in-flight GPT-5.4 stacks

$30
in / Mtok
$180
out / Mtok
1.1M
context
tok/s
Google · Flash

Gemini 2.5 Flash

6.7
AI Panel
Value 7.0
GA2025Cost-OptimizedMultimodalLong-Context

Best for mature mid-tier with thinking toggle

$0.3
in / Mtok
$2.5
out / Mtok
1.0M
context
222
tok/s
xAI · Coder

Grok Build 0.1

6.7
AI Panel
Value 7.5
GA2026Coding

Best for local-first coding agent for IP-sensitive teams

$1
in / Mtok
$2
out / Mtok
256K
context
tok/s
Meta · Large

Llama 3.1 70B

6.7
AI Panel
Value 7.5
GA2024Open-weightsOpen-WeightsCost-Optimized

Best for 70B base checkpoint for continued pretraining

$0.4
in / Mtok
$0.59
out / Mtok
128K
context
82
tok/s
OpenAI · Pro

GPT-5

6.5
AI Panel
Value 7.0
GA2025ReasoningCodingMultimodal

Best for legacy flagship on a migration path

$1.25
in / Mtok
$10
out / Mtok
400K
context
90
tok/s
OpenAI · nano

GPT-5 nano

6.4
AI Panel
Value 9.0
GA2025Cost-OptimizedEdge

Best for OpenAI cost floor for plumbing-grade work

$0.05
in / Mtok
$0.4
out / Mtok
400K
context
220
tok/s
Meta · Large

Llama 3.1 405B

6.3
AI Panel
Value 4.5
GA2024Open-weightsOpen-Weights

Best for largest open base for from-scratch fine-tuning

$3
in / Mtok
$3
out / Mtok
128K
context
29
tok/s
OpenAI · mini

GPT-5 mini

5.8
AI Panel
Value 6.0
GA2025Cost-OptimizedMultimodal

Best for superseded legacy mini, migrate to GPT-5.4 nano

$0.25
in / Mtok
$2
out / Mtok
400K
context
150
tok/s
Anthropic · Opus

Claude Opus 4.1

5.5
AI Panel
Value 2.5
GA2025ReasoningCodingMultimodal

Best for legacy snapshot, maintenance only

$15
in / Mtok
$75
out / Mtok
200K
context
45
tok/s