Back to AI Coding

AI Model Ranking

AI Coding Model Rankings

Top 50 from Artificial Analysis.

Snapshot: 2026-11-18

Source Artificial AnalysisMethodology

Metrics follow a snapshot of the Artificial Analysis public LLM leaderboard: Intelligence Index, cost per task, and speed reflect benchmark performance at the snapshot date. Provider pricing, availability, and rankings change frequently, and a high benchmark score does not guarantee the best result for a specific codebase or workflow.

Read the full methodology guide

AA Index top 50

65
62
60
59
57
55
54
51
50
50
46
44
43
36
36
35
35
33
32
31
30
30
29
29
29
28
28
27
26
25
24
23
20
20
19
19
18
18
14
14
14
12
12
12
9
8
6
6
6
6

Top 50 models

01

Current reasoning model

Anthropic - Claude Fable - 5.2 (max)

Anthropic Claude Fable 5.2 (max) leads the 2026-11-18 snapshot with Intelligence Index 65, 1M context, $26.54 cost per task, 112 tok/s median output, and 214.95s to first chunk. First chunk takes 214.95s, so it fits batch or async work better than interactive loops.

Best for: Maximum-effort Anthropic work on hard refactors and architecture decisions, at $26.54 per task with 214.95s to first chunk.

Anthropic

Context
1M
AA Index
65
Cost per task
$26.54
Speed
112 tok/s
First chunk
214.95s
Total response
219.41s
02

Current reasoning model

OpenAI - GPT Astra - 6.2 (max)

OpenAI GPT Astra 6.2 (max) ranks #2 in the 2026-11-18 snapshot with Intelligence Index 62, 1M context, $4.00 cost per task, 64 tok/s median output, and 147.65s to first chunk. First chunk takes 147.65s, so it fits batch or async work better than interactive loops.

Best for: Maximum-effort OpenAI work on hard refactors and architecture decisions, at $4.00 per task with 147.65s to first chunk.

OpenAI

Context
1M
AA Index
62
Cost per task
$4.00
Speed
64 tok/s
First chunk
147.65s
Total response
155.48s
03

Current reasoning model

Google - Gemini - 4 Carbon (high)

Google Gemini 4 Carbon (high) ranks #3 in the 2026-11-18 snapshot with Intelligence Index 60, 1M context, $5.48 cost per task, 62 tok/s median output, and 73.37s to first chunk. First chunk takes 73.37s, so it fits batch or async work better than interactive loops.

Best for: High-effort Google coding and multi-step agent runs, at $5.48 per task with 73.37s to first chunk.

Google

Context
1M
AA Index
60
Cost per task
$5.48
Speed
62 tok/s
First chunk
73.37s
Total response
81.47s
04

Current reasoning model

Anthropic - Claude Opus - 5.5 (max)

Anthropic Claude Opus 5.5 (max) ranks #4 in the 2026-11-18 snapshot with Intelligence Index 59, 1M context, $10.04 cost per task, 118 tok/s median output, and 355.12s to first chunk. First chunk takes 355.12s, so it fits batch or async work better than interactive loops.

Best for: Maximum-effort Anthropic work on hard refactors and architecture decisions, at $10.04 per task with 355.12s to first chunk.

Anthropic

Context
1M
AA Index
59
Cost per task
$10.04
Speed
118 tok/s
First chunk
355.12s
Total response
359.37s
05

Current reasoning model

OpenAI - GPT Sol - 6.2 (max)

OpenAI GPT Sol 6.2 (max) ranks #5 in the 2026-11-18 snapshot with Intelligence Index 57, 1M context, $1.01 cost per task, 85 tok/s median output, and 61.56s to first chunk. First chunk takes 61.56s, so it fits batch or async work better than interactive loops.

Best for: Maximum-effort OpenAI work on hard refactors and architecture decisions, at $1.01 per task with 61.56s to first chunk.

OpenAI

Context
1M
AA Index
57
Cost per task
$1.01
Speed
85 tok/s
First chunk
61.56s
Total response
67.43s
06

Current reasoning model

Anthropic - Claude Sonnet - 5.5 (max)

Anthropic Claude Sonnet 5.5 (max) ranks #6 in the 2026-11-18 snapshot with Intelligence Index 55, 1M context, $10.02 cost per task, 147 tok/s median output, and 290.29s to first chunk. First chunk takes 290.29s, so it fits batch or async work better than interactive loops.

Best for: Maximum-effort Anthropic work on hard refactors and architecture decisions, at $10.02 per task with 290.29s to first chunk.

Anthropic

Context
1M
AA Index
55
Cost per task
$10.02
Speed
147 tok/s
First chunk
290.29s
Total response
293.69s
07

Current reasoning model

Google - Gemini - 4 Hydrogen (xhigh)

Google Gemini 4 Hydrogen (xhigh) ranks #7 in the 2026-11-18 snapshot with Intelligence Index 54, 1M context, $3.16 cost per task, 174 tok/s median output, and 42.69s to first chunk.

Best for: Very high-effort Google work where extra reasoning is worth a longer wait, at $3.16 per task with 42.69s to first chunk.

Google

Context
1M
AA Index
54
Cost per task
$3.16
Speed
174 tok/s
First chunk
42.69s
Total response
45.57s
08

Current reasoning model

OpenAI - GPT Terra - 6.1 (max)

OpenAI GPT Terra 6.1 (max) ranks #8 in the 2026-11-18 snapshot with Intelligence Index 51, 1M context, $0.96 cost per task, 70 tok/s median output, and 144.41s to first chunk. First chunk takes 144.41s, so it fits batch or async work better than interactive loops.

Best for: Maximum-effort OpenAI work on hard refactors and architecture decisions, at $0.96 per task with 144.41s to first chunk.

OpenAI

Context
1M
AA Index
51
Cost per task
$0.96
Speed
70 tok/s
First chunk
144.41s
Total response
151.52s
09

Current reasoning model

Google - Gemini - 4 Argon (high)

Google Gemini 4 Argon (high) ranks #9 in the 2026-11-18 snapshot with Intelligence Index 50, 1M context, $4.97 cost per task, 67 tok/s median output, and 47.00s to first chunk. It is one of 2 models at Intelligence Index 50.

Best for: High-effort Google coding and multi-step agent runs, at $4.97 per task with 47.00s to first chunk.

Google

Context
1M
AA Index
50
Cost per task
$4.97
Speed
67 tok/s
First chunk
47.00s
Total response
54.51s
10

Current model; partial public metrics

Meta - Muse Wonder Full - Max

Meta Muse Wonder Full Max ranks #10 in the 2026-11-18 snapshot with Intelligence Index 50 and 1M context. Cost per task, output speed, and latency are not published yet. It is one of 2 models at Intelligence Index 50.

Best for: Maximum-effort Meta work on hard refactors and architecture decisions; cost, speed, and latency are not published yet.

Meta

Context
1M
AA Index
50
Cost per task
—
Speed
— tok/s
First chunk
—
Total response
—
11

Current reasoning model

Kimi - Kimi - K3.1 Pro

Kimi Kimi K3.1 Pro ranks #11 in the 2026-11-18 snapshot with Intelligence Index 46, 1M context, $3.68 cost per task, 31 tok/s median output, and 155.21s to first chunk. First chunk takes 155.21s, so it fits batch or async work better than interactive loops.

Best for: Kimi's flagship Kimi model for demanding coding work, at $3.68 per task with 155.21s to first chunk.

Kimi

Context
1M
AA Index
46
Cost per task
$3.68
Speed
31 tok/s
First chunk
155.21s
Total response
171.11s
12

Current reasoning model

Meta - Muse Wonder - Max

Meta Muse Wonder Max ranks #12 in the 2026-11-18 snapshot with Intelligence Index 44, 1.05M context, $6.95 cost per task, 120 tok/s median output, and 19.58s to first chunk.

Best for: Maximum-effort Meta work on hard refactors and architecture decisions, at $6.95 per task with 19.58s to first chunk.

Meta

Context
1.05M
AA Index
44
Cost per task
$6.95
Speed
120 tok/s
First chunk
19.58s
Total response
23.76s
13

Current reasoning model

Alibaba - Qwen - 4 Max

Alibaba Qwen 4 Max ranks #13 in the 2026-11-18 snapshot with Intelligence Index 43, 984k context, $3.27 cost per task, 53 tok/s median output, and 143.51s to first chunk. First chunk takes 143.51s, so it fits batch or async work better than interactive loops.

Best for: Maximum-effort Alibaba work on hard refactors and architecture decisions, at $3.27 per task with 143.51s to first chunk.

Alibaba

Context
984k
AA Index
43
Cost per task
$3.27
Speed
53 tok/s
First chunk
143.51s
Total response
152.98s
14

Current reasoning model

Anthropic - Claude Haiku - 5.5 (max)

Anthropic Claude Haiku 5.5 (max) ranks #14 in the 2026-11-18 snapshot with Intelligence Index 36, 1M context, $0.41 cost per task, 255 tok/s median output, and 190.13s to first chunk. It is one of 2 models at Intelligence Index 36. First chunk takes 190.13s, so it fits batch or async work better than interactive loops.

Best for: Maximum-effort Anthropic work on hard refactors and architecture decisions, at $0.41 per task with 190.13s to first chunk.

Anthropic

Context
1M
AA Index
36
Cost per task
$0.41
Speed
255 tok/s
First chunk
190.13s
Total response
192.09s
15

Current reasoning model

Meta - Muse Ember - Max

Meta Muse Ember Max ranks #15 in the 2026-11-18 snapshot with Intelligence Index 36, 1.05M context, $0.14 cost per task, 452 tok/s median output, and 4.85s to first chunk. It is one of 2 models at Intelligence Index 36. First chunk arrives in 4.85s, so it suits interactive loops.

Best for: Maximum-effort Meta work on hard refactors and architecture decisions, at $0.14 per task with 4.85s to first chunk.

Meta

Context
1.05M
AA Index
36
Cost per task
$0.14
Speed
452 tok/s
First chunk
4.85s
Total response
5.96s
16

Current reasoning model

SpaceXAI - Grok - 4.7 (xhigh)

SpaceXAI Grok 4.7 (xhigh) ranks #16 in the 2026-11-18 snapshot with Intelligence Index 35, 500k context, $7.77 cost per task, 72 tok/s median output, and 47.29s to first chunk. It is one of 2 models at Intelligence Index 35.

Best for: Very high-effort SpaceXAI work where extra reasoning is worth a longer wait, at $7.77 per task with 47.29s to first chunk.

SpaceXAI

Context
500k
AA Index
35
Cost per task
$7.77
Speed
72 tok/s
First chunk
47.29s
Total response
54.23s
17

Current reasoning model

Tencent - Hy - 4

Tencent Hy 4 ranks #17 in the 2026-11-18 snapshot with Intelligence Index 35, 1.05M context, $1.03 cost per task, 96 tok/s median output, and 111.73s to first chunk. It is one of 2 models at Intelligence Index 35. First chunk takes 111.73s, so it fits batch or async work better than interactive loops.

Best for: Tencent's flagship Hy model for demanding coding work, at $1.03 per task with 111.73s to first chunk.

Tencent

Context
1.05M
AA Index
35
Cost per task
$1.03
Speed
96 tok/s
First chunk
111.73s
Total response
116.94s
18

Current reasoning model

DeepSeek - DeepSeek - V4.1 Flash 1017 (max)

DeepSeek DeepSeek V4.1 Flash 1017 (max) ranks #18 in the 2026-11-18 snapshot with Intelligence Index 33, 1M context, $0.60 cost per task, 178 tok/s median output, and 37.93s to first chunk.

Best for: Maximum-effort DeepSeek work on hard refactors and architecture decisions, at $0.60 per task with 37.93s to first chunk.

DeepSeek

Context
1M
AA Index
33
Cost per task
$0.60
Speed
178 tok/s
First chunk
37.93s
Total response
40.74s
19

Current reasoning model

Z AI - GLM - 5.3 (max)

Z AI GLM 5.3 (max) ranks #19 in the 2026-11-18 snapshot with Intelligence Index 32, 1M context, $5.01 cost per task, 91 tok/s median output, and 69.29s to first chunk. First chunk takes 69.29s, so it fits batch or async work better than interactive loops.

Best for: Maximum-effort Z AI work on hard refactors and architecture decisions, at $5.01 per task with 69.29s to first chunk.

Z AI

Context
1M
AA Index
32
Cost per task
$5.01
Speed
91 tok/s
First chunk
69.29s
Total response
74.75s
20

Current reasoning model

Thinking Machines - Inkling - 2 Small

Thinking Machines Inkling 2 Small ranks #20 in the 2026-11-18 snapshot with Intelligence Index 31, 1.02M context, $1.18 cost per task, 119 tok/s median output, and 39.18s to first chunk.

Best for: Thinking Machines' flagship Inkling model for demanding coding work, at $1.18 per task with 39.18s to first chunk.

Thinking Machines

Context
1.02M
AA Index
31
Cost per task
$1.18
Speed
119 tok/s
First chunk
39.18s
Total response
43.37s
21

Current reasoning model

Xiaomi - MiMo - V2.6 Pro

Xiaomi MiMo V2.6 Pro ranks #21 in the 2026-11-18 snapshot with Intelligence Index 30, 1M context, $0.26 cost per task, 58 tok/s median output, and 25.73s to first chunk. It is one of 2 models at Intelligence Index 30.

Best for: Xiaomi's flagship MiMo model for demanding coding work, at $0.26 per task with 25.73s to first chunk.

Xiaomi

Context
1M
AA Index
30
Cost per task
$0.26
Speed
58 tok/s
First chunk
25.73s
Total response
34.35s
22

Current model; partial public metrics

InclusionAI - Ling - 3.1 Flash 1019

InclusionAI Ling 3.1 Flash 1019 ranks #22 in the 2026-11-18 snapshot with Intelligence Index 30, 1.05M context, and $3.19 cost per task. Output speed and latency are not published yet. It is one of 2 models at Intelligence Index 30.

Best for: High-volume InclusionAI inference where speed matters more than the top index; output speed and latency are not published yet.

InclusionAI

Context
1.05M
AA Index
30
Cost per task
$3.19
Speed
— tok/s
First chunk
—
Total response
—
23

Current reasoning model

MiniMax - MiniMax - M3.1 Flash

MiniMax MiniMax M3.1 Flash ranks #23 in the 2026-11-18 snapshot with Intelligence Index 29, 1M context, $0.49 cost per task, 131 tok/s median output, and 48.55s to first chunk. It is one of 3 models at Intelligence Index 29.

Best for: High-volume MiniMax inference where speed matters more than the top index, at $0.49 per task with 48.55s to first chunk.

MiniMax

Context
1M
AA Index
29
Cost per task
$0.49
Speed
131 tok/s
First chunk
48.55s
Total response
52.36s
24

Current reasoning model

Alibaba - Qwen - 3.8 Flash Next

Alibaba Qwen 3.8 Flash Next ranks #24 in the 2026-11-18 snapshot with Intelligence Index 29, 256k context, $0.57 cost per task, 84 tok/s median output, and 62.69s to first chunk. It is one of 3 models at Intelligence Index 29. First chunk takes 62.69s, so it fits batch or async work better than interactive loops.

Best for: High-volume Alibaba inference where speed matters more than the top index, at $0.57 per task with 62.69s to first chunk.

Alibaba

Context
256k
AA Index
29
Cost per task
$0.57
Speed
84 tok/s
First chunk
62.69s
Total response
68.66s
25

Current reasoning model

StepFun - Step - 5 Preview

StepFun Step 5 Preview ranks #25 in the 2026-11-18 snapshot with Intelligence Index 29, 1M context, $2.35 cost per task, 144 tok/s median output, and 21.77s to first chunk. It is one of 3 models at Intelligence Index 29.

Best for: StepFun's flagship Step model for demanding coding work, at $2.35 per task with 21.77s to first chunk.

StepFun

Context
1M
AA Index
29
Cost per task
$2.35
Speed
144 tok/s
First chunk
21.77s
Total response
25.25s
26

Current reasoning model

OpenAI - GPT Luna - 6 (max)

OpenAI GPT Luna 6 (max) ranks #26 in the 2026-11-18 snapshot with Intelligence Index 28, 1M context, $0.15 cost per task, 104 tok/s median output, and 64.04s to first chunk. It is one of 2 models at Intelligence Index 28. First chunk takes 64.04s, so it fits batch or async work better than interactive loops.

Best for: Maximum-effort OpenAI work on hard refactors and architecture decisions, at $0.15 per task with 64.04s to first chunk.

OpenAI

Context
1M
AA Index
28
Cost per task
$0.15
Speed
104 tok/s
First chunk
64.04s
Total response
68.86s
27

Current reasoning model

Z AI - GLM - 5.3 Flash

Z AI GLM 5.3 Flash ranks #27 in the 2026-11-18 snapshot with Intelligence Index 28, 1M context, $0.58 cost per task, 58 tok/s median output, and 84.58s to first chunk. It is one of 2 models at Intelligence Index 28. First chunk takes 84.58s, so it fits batch or async work better than interactive loops.

Best for: High-volume Z AI inference where speed matters more than the top index, at $0.58 per task with 84.58s to first chunk.

Z AI

Context
1M
AA Index
28
Cost per task
$0.58
Speed
58 tok/s
First chunk
84.58s
Total response
93.14s
28

Current reasoning model

Xiaomi - MiMo - V2.6 Flash

Xiaomi MiMo V2.6 Flash ranks #28 in the 2026-11-18 snapshot with Intelligence Index 27, 1M context, $0.13 cost per task, 40 tok/s median output, and 33.03s to first chunk.

Best for: High-volume Xiaomi inference where speed matters more than the top index, at $0.13 per task with 33.03s to first chunk.

Xiaomi

Context
1M
AA Index
27
Cost per task
$0.13
Speed
40 tok/s
First chunk
33.03s
Total response
45.47s
29

Current reasoning model

Mistral - Mistral Large - 4 (high)

Mistral Mistral Large 4 (high) ranks #29 in the 2026-11-18 snapshot with Intelligence Index 26, 1M context, $2.06 cost per task, 73 tok/s median output, and 64.46s to first chunk. First chunk takes 64.46s, so it fits batch or async work better than interactive loops.

Best for: High-effort Mistral coding and multi-step agent runs, at $2.06 per task with 64.46s to first chunk.

Mistral

Context
1M
AA Index
26
Cost per task
$2.06
Speed
73 tok/s
First chunk
64.46s
Total response
71.33s
30

Current reasoning model

Meta - Muse Spark Open - Xhigh

Meta Muse Spark Open Xhigh ranks #30 in the 2026-11-18 snapshot with Intelligence Index 25, 1.05M context, $2.02 cost per task, 190 tok/s median output, and 10.44s to first chunk.

Best for: Very high-effort Meta work where extra reasoning is worth a longer wait, at $2.02 per task with 10.44s to first chunk.

Meta

Context
1.05M
AA Index
25
Cost per task
$2.02
Speed
190 tok/s
First chunk
10.44s
Total response
13.08s
31

Current reasoning model

Sapiens AI - Agnes - 2.5 Pro Theta

Sapiens AI Agnes 2.5 Pro Theta ranks #31 in the 2026-11-18 snapshot with Intelligence Index 24, 1M context, $0.57 cost per task, 203 tok/s median output, and 8.10s to first chunk.

Best for: Sapiens AI's flagship Agnes model for demanding coding work, at $0.57 per task with 8.10s to first chunk.

Sapiens AI

Context
1M
AA Index
24
Cost per task
$0.57
Speed
203 tok/s
First chunk
8.10s
Total response
10.57s
32

Current reasoning model

DeepSeek - DeepSeek - V4 Pro 0813 (max)

DeepSeek DeepSeek V4 Pro 0813 (max) ranks #32 in the 2026-11-18 snapshot with Intelligence Index 23, 1M context, $1.77 cost per task, 97 tok/s median output, and 54.34s to first chunk.

Best for: Maximum-effort DeepSeek work on hard refactors and architecture decisions, at $1.77 per task with 54.34s to first chunk.

DeepSeek

Context
1M
AA Index
23
Cost per task
$1.77
Speed
97 tok/s
First chunk
54.34s
Total response
59.48s
33

Current model; partial public metrics

Nanbeige - Nanbeige - 4.3-33B

Nanbeige Nanbeige 4.3-33B ranks #33 in the 2026-11-18 snapshot with Intelligence Index 20 and 524k context. Cost per task, output speed, and latency are not published yet. It is one of 2 models at Intelligence Index 20.

Best for: Nanbeige's flagship Nanbeige model for demanding coding work; cost, speed, and latency are not published yet.

Nanbeige

Context
524k
AA Index
20
Cost per task
—
Speed
— tok/s
First chunk
—
Total response
—
34

Current reasoning model

Alibaba - Qwen - 3.8 27B (xhigh)

Alibaba Qwen 3.8 27B (xhigh) ranks #34 in the 2026-11-18 snapshot with Intelligence Index 20, 256k context, $3.53 cost per task, 48 tok/s median output, and 121.79s to first chunk. It is one of 2 models at Intelligence Index 20. First chunk takes 121.79s, so it fits batch or async work better than interactive loops.

Best for: Very high-effort Alibaba work where extra reasoning is worth a longer wait, at $3.53 per task with 121.79s to first chunk.

Alibaba

Context
256k
AA Index
20
Cost per task
$3.53
Speed
48 tok/s
First chunk
121.79s
Total response
132.19s
35

Current model; partial public metrics

China Mobile - JT - 4.1 Flash 236B A21B

China Mobile JT 4.1 Flash 236B A21B ranks #35 in the 2026-11-18 snapshot with Intelligence Index 19 and 262k context. Cost per task, output speed, and latency are not published yet. It is one of 2 models at Intelligence Index 19.

Best for: High-volume China Mobile inference where speed matters more than the top index; cost, speed, and latency are not published yet.

China Mobile

Context
262k
AA Index
19
Cost per task
—
Speed
— tok/s
First chunk
—
Total response
—
36

Current reasoning model

MiniMax - MiniMax - M3

MiniMax MiniMax M3 ranks #36 in the 2026-11-18 snapshot with Intelligence Index 19, 1M context, $1.29 cost per task, 108 tok/s median output, and 69.98s to first chunk. It is one of 2 models at Intelligence Index 19. First chunk takes 69.98s, so it fits batch or async work better than interactive loops.

Best for: MiniMax's flagship MiniMax model for demanding coding work, at $1.29 per task with 69.98s to first chunk.

MiniMax

Context
1M
AA Index
19
Cost per task
$1.29
Speed
108 tok/s
First chunk
69.98s
Total response
74.62s
37

Current reasoning model

Institute of Foundation Models - K2 Horizon - 375B A23B

Institute of Foundation Models K2 Horizon 375B A23B ranks #37 in the 2026-11-18 snapshot with Intelligence Index 18, 524k context, $0.00 cost per task, 221 tok/s median output, and 19.59s to first chunk. It is one of 2 models at Intelligence Index 18.

Best for: Institute of Foundation Models' flagship K2 Horizon model for demanding coding work, at $0.00 per task with 19.59s to first chunk.

Institute of Foundation Models

Context
524k
AA Index
18
Cost per task
$0.00
Speed
221 tok/s
First chunk
19.59s
Total response
21.85s
38

Current reasoning model

Reflection - Beam - 1.0

Reflection Beam 1.0 ranks #38 in the 2026-11-18 snapshot with Intelligence Index 18, 1.05M context, $0.28 cost per task, 263 tok/s median output, and 7.30s to first chunk. It is one of 2 models at Intelligence Index 18.

Best for: Reflection's flagship Beam model for demanding coding work, at $0.28 per task with 7.30s to first chunk.

Reflection

Context
1.05M
AA Index
18
Cost per task
$0.28
Speed
263 tok/s
First chunk
7.30s
Total response
9.20s
39

Current reasoning model

Upstage - Solar Pro - 4

Upstage Solar Pro 4 ranks #39 in the 2026-11-18 snapshot with Intelligence Index 14, 512k context, $0.96 cost per task, 72 tok/s median output, and 31.47s to first chunk. It is one of 3 models at Intelligence Index 14.

Best for: Upstage's flagship Solar Pro model for demanding coding work, at $0.96 per task with 31.47s to first chunk.

Upstage

Context
512k
AA Index
14
Cost per task
$0.96
Speed
72 tok/s
First chunk
31.47s
Total response
38.41s
40

Current model; partial public metrics

Apodex - Apodex - 1.1

Apodex Apodex 1.1 ranks #40 in the 2026-11-18 snapshot with Intelligence Index 14, 256k context, and $0.76 cost per task. Output speed and latency are not published yet. It is one of 3 models at Intelligence Index 14.

Best for: Apodex's flagship Apodex model for demanding coding work; output speed and latency are not published yet.

Apodex

Context
256k
AA Index
14
Cost per task
$0.76
Speed
— tok/s
First chunk
—
Total response
—
41

Current reasoning model

NVIDIA - Nemotron - 3.5 Super VL

NVIDIA Nemotron 3.5 Super VL ranks #41 in the 2026-11-18 snapshot with Intelligence Index 14, 1M context, $0.48 cost per task, 254 tok/s median output, and 13.69s to first chunk. It is one of 3 models at Intelligence Index 14.

Best for: NVIDIA's flagship Nemotron model for demanding coding work, at $0.48 per task with 13.69s to first chunk.

NVIDIA

Context
1M
AA Index
14
Cost per task
$0.48
Speed
254 tok/s
First chunk
13.69s
Total response
15.66s
42

Current reasoning model

Thinking Machines - Inkling - Xhigh

Thinking Machines Inkling Xhigh ranks #42 in the 2026-11-18 snapshot with Intelligence Index 12, 1M context, $1.35 cost per task, 105 tok/s median output, and 14.63s to first chunk. It is one of 3 models at Intelligence Index 12.

Best for: Very high-effort Thinking Machines work where extra reasoning is worth a longer wait, at $1.35 per task with 14.63s to first chunk.

Thinking Machines

Context
1M
AA Index
12
Cost per task
$1.35
Speed
105 tok/s
First chunk
14.63s
Total response
19.40s
43

Current reasoning model

NVIDIA - Nemotron - 3 Ultra 550B A55B (Reasoning)

NVIDIA Nemotron 3 Ultra 550B A55B (Reasoning) ranks #43 in the 2026-11-18 snapshot with Intelligence Index 12, 262k context, $2.21 cost per task, 195 tok/s median output, and 3.67s to first chunk. It is one of 3 models at Intelligence Index 12. First chunk arrives in 3.67s, so it suits interactive loops.

Best for: High-effort NVIDIA coding and multi-step agent runs, at $2.21 per task with 3.67s to first chunk.

NVIDIA

Context
262k
AA Index
12
Cost per task
$2.21
Speed
195 tok/s
First chunk
3.67s
Total response
6.23s
44

Current reasoning model

Google - Gemini - 3.5 Flash-Lite

Google Gemini 3.5 Flash-Lite ranks #44 in the 2026-11-18 snapshot with Intelligence Index 12, 1M context, $0.69 cost per task, 329 tok/s median output, and 4.82s to first chunk. It is one of 3 models at Intelligence Index 12. First chunk arrives in 4.82s, so it suits interactive loops.

Best for: High-volume Google inference where speed matters more than the top index, at $0.69 per task with 4.82s to first chunk.

Google

Context
1M
AA Index
12
Cost per task
$0.69
Speed
329 tok/s
First chunk
4.82s
Total response
6.34s
45

Current model; partial public metrics

SK Telecom - A.X - K2

SK Telecom A.X K2 ranks #45 in the 2026-11-18 snapshot with Intelligence Index 9 and 262k context. Cost per task, output speed, and latency are not published yet.

Best for: SK Telecom's flagship A.X model for demanding coding work; cost, speed, and latency are not published yet.

SK Telecom

Context
262k
AA Index
9
Cost per task
—
Speed
— tok/s
First chunk
—
Total response
—
46

Current model; partial public metrics

LG AI Research - K-EXAONE - 2.0

LG AI Research K-EXAONE 2.0 ranks #46 in the 2026-11-18 snapshot with Intelligence Index 8 and 262k context. Cost per task, output speed, and latency are not published yet.

Best for: LG AI Research's flagship K-EXAONE model for demanding coding work; cost, speed, and latency are not published yet.

LG AI Research

Context
262k
AA Index
8
Cost per task
—
Speed
— tok/s
First chunk
—
Total response
—
47

Current reasoning model

Meta - Muse Glimmer - High

Meta Muse Glimmer High ranks #47 in the 2026-11-18 snapshot with Intelligence Index 6, 131k context, $0.18 cost per task, 93 tok/s median output, and 19.29s to first chunk. It is one of 4 models at Intelligence Index 6.

Best for: High-effort Meta coding and multi-step agent runs, at $0.18 per task with 19.29s to first chunk.

Meta

Context
131k
AA Index
6
Cost per task
$0.18
Speed
93 tok/s
First chunk
19.29s
Total response
24.67s
48

Current reasoning model

Cohere - Command - A+

Cohere Command A+ ranks #48 in the 2026-11-18 snapshot with Intelligence Index 6, 192k context, $0.00 cost per task, 167 tok/s median output, and 29.84s to first chunk. It is one of 4 models at Intelligence Index 6.

Best for: Cohere's flagship Command model for demanding coding work, at $0.00 per task with 29.84s to first chunk.

Cohere

Context
192k
AA Index
6
Cost per task
$0.00
Speed
167 tok/s
First chunk
29.84s
Total response
32.84s
49

Current model; partial public metrics

Liquid AI - LFM - 3-1.7B

Liquid AI LFM 3-1.7B ranks #49 in the 2026-11-18 snapshot with Intelligence Index 6, 262k context, and $0.17 cost per task. Output speed and latency are not published yet. It is one of 4 models at Intelligence Index 6.

Best for: Liquid AI's flagship LFM model for demanding coding work; output speed and latency are not published yet.

Liquid AI

Context
262k
AA Index
6
Cost per task
$0.17
Speed
— tok/s
First chunk
—
Total response
—
50

Current reasoning model

NVIDIA - Nemotron - 3.5 Lightning

NVIDIA Nemotron 3.5 Lightning ranks #50 in the 2026-11-18 snapshot with Intelligence Index 6, 1M context, $0.31 cost per task, 277 tok/s median output, and 13.95s to first chunk. It is one of 4 models at Intelligence Index 6.

Best for: NVIDIA's flagship Nemotron model for demanding coding work, at $0.31 per task with 13.95s to first chunk.

NVIDIA

Context
1M
AA Index
6
Cost per task
$0.31
Speed
277 tok/s
First chunk
13.95s
Total response
15.76s