Workbench

Models workbench

Filter the catalog, read the price–performance frontier, and carry a two-to-four model selection into a shareable comparison. Values with no source measurement stay unavailable.

Price–performance frontier

The line joins non-dominated models using a disclosed 75% input / 25% output token blend. Cost is on a logarithmic axis by default, because token prices span four orders of magnitude; untick it for a linear axis.

Frontier only$ per successful taskLog cost scale

Model catalog

340 of 340 source-backed models.Pick two to four to compare; the tray follows you down the page and the URL updates as you work.

Evidence key: ObservedLast goodUnavailable

Claude Opus 5.5

Anthropic · rank 1

Proprietary

Context1M tokens

Input / 1M$4.00

Output / 1M$20.00

Runtime

Overall86.68

TTFT573.15 s

Throughput95 tok/s

Capability

Agentic65.3

Coding87.5

Knowledge85.4

Reasoning86.1

Multimodal & Grounded92.2

Instruction FollowingUnavailable

Math97.3

6 of 7 axes scoredFull evidence

GPT-6.1 Sol

OpenAI · rank 2

Proprietary

Context1.05M tokens

Input / 1M$2.00

Output / 1M$10.00

Runtime

Overall84.22

TTFTUnavailable

Throughput38.5 tok/s

Capability

AgenticUnavailable

Coding74.1

Knowledge83.8

Reasoning90.5

Multimodal & Grounded87.3

Instruction FollowingUnavailable

Math97.3

5 of 7 axes scoredFull evidence

GPT-6 Astra

OpenAI · rank 3

Proprietary

Context1.05M tokens

Input / 1M$10.00

Output / 1M$50.00

Runtime

Overall84.04

TTFT265.1 s

Throughput66 tok/s

Capability

Agentic86.5

Coding76.4

Knowledge84.3

Reasoning93.3

Multimodal & Grounded87.3

Instruction FollowingUnavailable

Math97.3

6 of 7 axes scoredFull evidence

Claude Fable 5.1

Anthropic · rank 4

Proprietary

Context1M tokens

Input / 1M$10.00

Output / 1M$50.00

Runtime

Overall83.22

TTFT206.2 s

Throughput69 tok/s

Capability

Agentic59.8

Coding91.2

Knowledge95.4

Reasoning85.2

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

Math96.8

5 of 7 axes scoredFull evidence

Gemini 4 Argon

Google · rank 5

Proprietary

Context1M tokens

Input / 1M$2.00

Output / 1M$10.00

Runtime

Overall82.89

TTFT47.97 s

Throughput83 tok/s

Capability

Agentic79.6

Coding74

Knowledge74.9

Reasoning86.9

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

Math92.8

5 of 7 axes scoredFull evidence

Claude Sonnet 5.5

Anthropic · rank 6

Proprietary

Context1M tokens

Input / 1M$2.00

Output / 1M$10.00

Runtime

Overall82.24

TTFT394.99 s

Throughput136 tok/s

Capability

Agentic78.8

Coding87.5

Knowledge83.4

Reasoning81.8

Multimodal & Grounded86.7

Instruction FollowingUnavailable

Math95.2

6 of 7 axes scoredFull evidence

Claude Fable 5

Anthropic · rank 7

Proprietary

Context1M tokens

Input / 1M$10.00

Output / 1M$50.00

Runtime

Overall81.33

TTFT100.13 s

Throughput68 tok/s

Capability

Agentic82.9

Coding90.3

Knowledge85

Reasoning84.3

Multimodal & Grounded66

Instruction Following77.1

Math96

7 of 7 axes scoredFull evidence

Claude Opus 5

Anthropic · rank 8

Proprietary

ContextUnavailable

Input / 1M$5.00

Output / 1M$25.00

Runtime

Overall80.07

TTFT59.73 s

Throughput62 tok/s

Capability

Agentic85.2

Coding90.4

Knowledge95.3

Reasoning81

Multimodal & Grounded92.2

Instruction FollowingUnavailable

Math91.6

6 of 7 axes scoredFull evidence

GPT-6 Sol

OpenAI · rank 9

Proprietary

Context1.05M tokens

Input / 1M$2.00

Output / 1M$10.00

Runtime

Overall78.6

TTFTUnavailable

Throughput64 tok/s

Capability

Agentic73.7

Coding77.5

Knowledge79.9

Reasoning73.1

Multimodal & Grounded86.6

Instruction FollowingUnavailable

Math97.3

6 of 7 axes scoredFull evidence

GPT-5.6 Sol

OpenAI · rank 10

Proprietary

Context1.05M tokens

Input / 1M$4.00

Output / 1M$20.00

Runtime

Overall75.29

TTFT100.66 s

Throughput94 tok/s

Capability

Agentic83.7

Coding64.7

Knowledge81.6

Reasoning76.4

Multimodal & Grounded89.8

Instruction Following89.1

Math95.8

7 of 7 axes scoredFull evidence

Muse Wonder Full

Meta · rank 11

Proprietary

Context1M tokens

Input / 1MUnavailable

Output / 1MUnavailable

Runtime

Overall74.75

TTFTUnavailable

ThroughputUnavailable

Capability

AgenticUnavailable

Coding69

Knowledge67.6

Reasoning82.5

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

MathUnavailable

3 of 7 axes scoredFull evidence

MiMo-V2.6-Pro

Xiaomi · rank 12

Open Weight

Context1M tokens

Input / 1M$0.44

Output / 1M$0.87

Runtime

Overall73.6

TTFT43.15 s

Throughput50 tok/s

Capability

Agentic77.8

Coding63.3

Knowledge58.4

Reasoning82.5

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

MathUnavailable

4 of 7 axes scoredFull evidence

GPT-5.5 Pro

OpenAI · rank 13

Proprietary

Context1M tokens

Input / 1M$30.00

Output / 1M$180.00

Runtime

Overall73.46

TTFTUnavailable

ThroughputUnavailable

Capability

Agentic80.4

CodingUnavailable

Knowledge82

Reasoning78.9

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

Math94

4 of 7 axes scoredFull evidence

Muse Wonder

Meta · rank 14

Proprietary

Context1M tokens

Input / 1M$4.00

Output / 1M$20.00

Runtime

Overall71.48

TTFT44.51 s

Throughput161 tok/s

Capability

AgenticUnavailable

CodingUnavailable

Knowledge65.3

Reasoning80.4

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

MathUnavailable

2 of 7 axes scoredFull evidence

GPT-5.6 Terra

OpenAI · rank 15

Proprietary

Context1.05M tokens

Input / 1M$2.00

Output / 1M$12.00

Runtime

Overall70.16

TTFTUnavailable

ThroughputUnavailable

Capability

Agentic76

Coding64.3

Knowledge73.7

Reasoning69.1

Multimodal & Grounded84.5

Instruction Following87.1

Math90.1

7 of 7 axes scoredFull evidence

Kimi K3

Moonshot AI · rank 16

Pending

Context1.05M tokens

Input / 1M$3.00

Output / 1M$15.00

Runtime

Overall69.95

TTFT51.11 s

Throughput42 tok/s

Capability

Agentic81.9

Coding74.6

Knowledge83.9

Reasoning69.8

Multimodal & Grounded92.6

Instruction FollowingUnavailable

MathUnavailable

5 of 7 axes scoredFull evidence

Claude Opus 4.8

Anthropic · rank 17

Proprietary

Context1M tokens

Input / 1M$5.00

Output / 1M$25.00

Runtime

Overall69.93

TTFT35.96 s

Throughput65 tok/s

Capability

Agentic74.1

Coding75.1

Knowledge87.5

Reasoning62.8

Multimodal & Grounded91.2

Instruction Following75.4

Math84.4

7 of 7 axes scoredFull evidence

Qwen3.8 Max

Alibaba · rank 18

Open Weight

Context1M tokens

Input / 1MUnavailable

Output / 1MUnavailable

Runtime

Overall69.73

TTFT53.6 s

Throughput40 tok/s

Capability

Agentic82.9

Coding70.6

Knowledge71.9

Reasoning87

Multimodal & Grounded91.2

Instruction Following94.7

Math78.9

7 of 7 axes scoredFull evidence

Step 5 Preview

StepFun · rank 19

Open Weight

Context1M tokens

Input / 1M$1.00

Output / 1M$2.70

Runtime

Overall69.72

TTFT24.09 s

Throughput93 tok/s

Capability

Agentic78.5

Coding77.6

Knowledge70.6

Reasoning82.5

Multimodal & Grounded74.6

Instruction FollowingUnavailable

MathUnavailable

5 of 7 axes scoredFull evidence

GPT-5.5

OpenAI · rank 20

Proprietary

Context1M tokens

Input / 1M$5.00

Output / 1M$30.00

Runtime

Overall69.18

TTFT37.05 s

Throughput98 tok/s

Capability

Agentic81.7

Coding55.5

Knowledge80.8

Reasoning70.3

Multimodal & Grounded75.5

Instruction Following93.3

Math91.2

7 of 7 axes scoredFull evidence

Gemini 3.8 Flash

Google · rank 21

Proprietary

Context1M tokens

Input / 1M$0.75

Output / 1M$3.75

Runtime

Overall69.09

TTFT26.25 s

Throughput149 tok/s

Capability

Agentic72.6

Coding77.5

Knowledge80.6

Reasoning74.3

Multimodal & Grounded87.3

Instruction FollowingUnavailable

Math69.7

6 of 7 axes scoredFull evidence

GPT-5.4 Pro

OpenAI · rank 22

Proprietary

Context1.05M tokens

Input / 1M$30.00

Output / 1M$180.00

Runtime

Overall68.69

TTFT151.79 s

Throughput74 tok/s

Capability

Agentic79.3

CodingUnavailable

Knowledge83.5

Reasoning78.1

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

Math86.5

4 of 7 axes scoredFull evidence

Claude Opus 4.7 (Adaptive)

Anthropic · rank 23

Proprietary

Context1M tokens

Input / 1M$5.00

Output / 1M$25.00

Runtime

Overall68.67

TTFTUnavailable

ThroughputUnavailable

Capability

Agentic71.8

Coding65.6

Knowledge82

Reasoning57.4

Multimodal & Grounded53.7

Instruction FollowingUnavailable

MathUnavailable

5 of 7 axes scoredFull evidence

Ling 3.1 Flash

InclusionAI · rank 24

Open Weight

Context262K tokens

Input / 1MUnavailable

Output / 1MUnavailable

Runtime

Overall68.05

TTFT11.16 s

Throughput215 tok/s

Capability

Agentic82.3

Coding69.1

Knowledge57.4

Reasoning82

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

MathUnavailable

4 of 7 axes scoredFull evidence

GLM-5.3

Z.AI · rank 25

Open WeightSelf-hosted

Context1M tokens

Input / 1MUnavailable

Output / 1MUnavailable

Runtime

Overall67.96

TTFT25.82 s

Throughput86 tok/s

Capability

Agentic74.5

Coding58.6

Knowledge64.8

Reasoning79.8

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

Math71.5

5 of 7 axes scoredFull evidence

Claude Haiku 5.5

Anthropic · rank 26

Proprietary

Context1M tokens

Input / 1M$0.10

Output / 1M$0.50

Runtime

Overall67.22

TTFT244.84 s

Throughput237 tok/s

Capability

AgenticUnavailable

Coding64.8

Knowledge67.3

Reasoning60.1

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

Math79.1

4 of 7 axes scoredFull evidence

MiniMax M3.1 Flash

MiniMax · rank 27

Proprietary

Context1M tokens

Input / 1M$0.15

Output / 1M$0.60

Runtime

Overall67.09

TTFT22.85 s

Throughput93 tok/s

Capability

Agentic73.7

CodingUnavailable

Knowledge74.2

Reasoning79.8

Multimodal & Grounded73.5

Instruction FollowingUnavailable

MathUnavailable

4 of 7 axes scoredFull evidence

Grok 4.6

xAI · rank 28

Proprietary

Context500K tokens

Input / 1M$2.00

Output / 1M$6.00

Runtime

Overall67.08

TTFT35.43 s

Throughput61 tok/s

Capability

Agentic77.6

Coding77.3

Knowledge75.9

Reasoning61.4

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

Math70.1

5 of 7 axes scoredFull evidence

Gemini 3.7 Flash

Google · rank 29

Proprietary

Context1M tokens

Input / 1M$0.75

Output / 1M$3.75

Runtime

Overall66.86

TTFT8.52 s

Throughput285 tok/s

Capability

Agentic64.7

Coding78.6

Knowledge80.7

Reasoning81.6

Multimodal & Grounded87.2

Instruction FollowingUnavailable

Math74.8

6 of 7 axes scoredFull evidence

GPT-5.4

OpenAI · rank 30

Proprietary

Context1.05M tokens

Input / 1M$2.50

Output / 1M$15.00

Runtime

Overall66.7

TTFT157.51 s

Throughput89 tok/s

Capability

Agentic80.3

Coding47.1

Knowledge77.3

Reasoning64.2

Multimodal & Grounded73.1

Instruction Following90.7

Math82

7 of 7 axes scoredFull evidence

GPT-5.2 Pro

OpenAI · rank 31

Proprietary

Context400K tokens

Input / 1M$21.00

Output / 1M$168.00

Runtime

Overall66.68

TTFTUnavailable

ThroughputUnavailable

Capability

AgenticUnavailable

CodingUnavailable

KnowledgeUnavailable

ReasoningUnavailable

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

MathUnavailable

0 of 7 axes scoredFull evidence

GPT-6 Luna

OpenAI · rank 32

Proprietary

Context1.05M tokens

Input / 1M$0.10

Output / 1M$0.50

Runtime

Overall66.51

TTFTUnavailable

Throughput58 tok/s

Capability

Agentic73.3

Coding66.3

Knowledge68.2

Reasoning59.2

Multimodal & Grounded81.9

Instruction FollowingUnavailable

Math83.7

6 of 7 axes scoredFull evidence

Muse Spark 1.1

Meta · rank 33

Proprietary

Context1M tokens

Input / 1MUnavailable

Output / 1MUnavailable

Runtime

Overall66.08

TTFT12.12 s

Throughput217 tok/s

Capability

Agentic66.9

Coding60.2

Knowledge90.8

Reasoning78.4

Multimodal & Grounded81.9

Instruction FollowingUnavailable

MathUnavailable

5 of 7 axes scoredFull evidence

Muse Spark 1.2

Meta · rank 34

Proprietary

Context1M tokens

Input / 1M$1.25

Output / 1M$4.25

Runtime

Overall65.88

TTFT28.37 s

Throughput154 tok/s

Capability

Agentic75.6

Coding62.9

Knowledge72.2

Reasoning79.3

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

MathUnavailable

4 of 7 axes scoredFull evidence

DeepSeek V4.1 Flash 1017

DeepSeek · rank 35

Open Weight

Context1M tokens

Input / 1M$0.30

Output / 1M$1.20

Runtime

Overall65.66

TTFT10.66 s

Throughput209 tok/s

Capability

Agentic83.5

Coding62.9

Knowledge69.7

Reasoning82.5

Multimodal & Grounded77.9

Instruction FollowingUnavailable

Math70.1

6 of 7 axes scoredFull evidence

Claude Opus 4.7

Anthropic · rank 36

Proprietary

Context1M tokens

Input / 1M$5.00

Output / 1M$25.00

Runtime

Overall65.4

TTFT18.77 s

Throughput58 tok/s

Capability

Agentic37.6

Coding69

Knowledge77.5

Reasoning31.4

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

Math72.9

5 of 7 axes scoredFull evidence

Claude Sonnet 5

Anthropic · rank 37

Proprietary

Context1M tokens

Input / 1M$2.00

Output / 1M$10.00

Runtime

Overall64.89

TTFT147.1 s

Throughput75 tok/s

Capability

Agentic77.5

Coding62.8

Knowledge81.2

Reasoning81.3

Multimodal & Grounded82

Instruction FollowingUnavailable

Math69.4

6 of 7 axes scoredFull evidence

Gemini 3.1 Pro

Google · rank 38

Proprietary

Context1M tokens

Input / 1M$2.00

Output / 1M$12.00

Runtime

Overall64.73

TTFT21.25 s

Throughput132 tok/s

Capability

Agentic69

Coding78.1

Knowledge77.1

Reasoning58.2

Multimodal & Grounded80.1

Instruction FollowingUnavailable

Math65

6 of 7 axes scoredFull evidence

MiMo-V2.6-Flash

Xiaomi · rank 39

Open Weight

Context1M tokens

Input / 1M$0.14

Output / 1M$0.28

Runtime

Overall64.57

TTFTUnavailable

ThroughputUnavailable

Capability

Agentic75.7

Coding64.1

Knowledge49.9

Reasoning76

Multimodal & Grounded72.2

Instruction FollowingUnavailable

MathUnavailable

5 of 7 axes scoredFull evidence

GPT-5.6 Luna

OpenAI · rank 40

Proprietary

Context1.05M tokens

Input / 1M$0.20

Output / 1M$1.20

Runtime

Overall64.39

TTFTUnavailable

ThroughputUnavailable

Capability

Agentic70.2

Coding56.4

Knowledge70.9

Reasoning58.2

Multimodal & Grounded79.6

Instruction FollowingUnavailable

Math84

6 of 7 axes scoredFull evidence

dots3-note Preview

Dots Studio · rank 41

Open WeightSelf-hosted

Context512K tokens

Input / 1MUnavailable

Output / 1MUnavailable

Runtime

Overall63.93

TTFTUnavailable

ThroughputUnavailable

Capability

Agentic71.1

Coding57.2

Knowledge78

Reasoning76.6

Multimodal & Grounded79.6

Instruction Following89.1

MathUnavailable

6 of 7 axes scoredFull evidence

Grok 4.5

xAI · rank 42

Proprietary

Context500K tokens

Input / 1M$2.00

Output / 1M$6.00

Runtime

Overall63.3

TTFT9.9 s

Throughput60 tok/s

Capability

Agentic74.3

Coding67.7

Knowledge77

Reasoning54.1

Multimodal & Grounded82.9

Instruction FollowingUnavailable

Math62.9

6 of 7 axes scoredFull evidence

DeepSeek V4 Pro 0813

DeepSeek · rank 43

Open Weight

Context1M tokens

Input / 1M$1.32

Output / 1M$3.96

Runtime

Overall63.21

TTFT1.68 s

Throughput101 tok/s

Capability

Agentic73.5

Coding59.4

Knowledge73.1

Reasoning60.7

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

Math87.4

5 of 7 axes scoredFull evidence

Sage-A1.5

InternScience · rank 44

Open WeightSelf-hosted

Context262K tokens

Input / 1MUnavailable

Output / 1MUnavailable

Runtime

Overall62.95

TTFTUnavailable

ThroughputUnavailable

Capability

Agentic69

Coding51.3

KnowledgeUnavailable

Reasoning80.6

Multimodal & GroundedUnavailable

Instruction Following91.9

MathUnavailable

4 of 7 axes scoredFull evidence

Qwen3.8-Flash-Next

Alibaba · rank 45

Open WeightSelf-hosted

Context262K tokens

Input / 1MUnavailable

Output / 1MUnavailable

Runtime

Overall62.83

TTFT38.3 s

Throughput55 tok/s

Capability

Agentic40

Coding58.6

Knowledge58.1

Reasoning79.8

Multimodal & Grounded87.7

Instruction Following91.2

MathUnavailable

6 of 7 axes scoredFull evidence

Gemini 3.6 Flash

Google · rank 46

Proprietary

Context1M tokens

Input / 1M$0.75

Output / 1M$3.75

Runtime

Overall62.79

TTFT14.25 s

Throughput215 tok/s

Capability

Agentic79.5

Coding60.6

Knowledge76.8

Reasoning67.6

Multimodal & Grounded87

Instruction FollowingUnavailable

Math63.5

6 of 7 axes scoredFull evidence

Muse Spark Open

Meta · rank 47

Open Weight

Context1M tokens

Input / 1M$1.25

Output / 1M$4.25

Runtime

Overall62.76

TTFT15.23 s

Throughput139 tok/s

Capability

AgenticUnavailable

CodingUnavailable

Knowledge63.3

Reasoning78.8

Multimodal & Grounded83.6

Instruction FollowingUnavailable

MathUnavailable

3 of 7 axes scoredFull evidence

Claude Opus 4.6 (Adaptive)

Anthropic · rank 48

Proprietary

Context1M tokens

Input / 1MUnavailable

Output / 1MUnavailable

Runtime

Overall62.72

TTFTUnavailable

ThroughputUnavailable

Capability

AgenticUnavailable

CodingUnavailable

KnowledgeUnavailable

Reasoning66.6

Multimodal & GroundedUnavailable

Instruction FollowingUnavailable

MathUnavailable

1 of 7 axes scoredFull evidence

76 of these rows are marked self-hosted by the source: open-weight models with no first-party hosted token rate. 12 of them show a hosted rate matched from another provider; the rest report an unavailable price rather than zero.

Time to first token: latency to the first answer chunk (Artificial Analysis, via BenchLM). Reasoning models include thinking time, so values can run to tens or hundreds of seconds.

Lifecycle & retirement watch

The next confirmed retirement notices returned by the lifecycle source.

No upcoming confirmed retirement.

Every confirmed notice the source returned already has a past effective date.

Open lifecycle radar

Release timeline

Most recent release dates from the catalog source.

2026-10-19Mercury 3Inception · Proprietary

2026-10-19MiniMax M3.1 FlashMiniMax · Proprietary

2026-10-19Spark-X2.6-4BSparkLLM · Open Weight

2026-10-17DeepSeek V4.1 Flash 1017DeepSeek · Open Weight

2026-10-16Perplexity Decider Flash v1.2 4BPerplexity · Open Weight

2026-10-16Perplexity Decider v1.2 27BPerplexity · Open Weight

2026-10-16Beam 1.0 501B-A23BReflection AI · Open Weight

2026-10-15d1-ultraLiquid AI · Proprietary

2026-10-15Muse Spark OpenMeta · Open Weight

2026-10-15Nemotron 3.5 Bolt 74B-A9BNVIDIA · Open Weight

2026-10-15Nemotron 3.5 Super VL 120B A12BNVIDIA · Open Weight

2026-10-15Sage-A1.5InternScience · Open Weight