VerdictPal
Join

Benchmarks / Performance

AA time to first token

TTFT

Median seconds until the first token returns on the model's default API provider with a ~1k-token input prompt.

Artificial AnalysisPerformanceFlagshipActiveUnknown contaminationSince 2026
What this does not measure
  • End-to-end task time for agent loops — TTFT is one hop, not a full coding session.
  • Non-default providers — AA pins one default route per model for comparability.
Analysis

Why this benchmark is useful

Time-to-first-token is the pause before anything appears — useful when snappy starts matter more than sustained generation speed.

Scope

Coverage map

Task family
Performance
Format
Single-shot API request with streaming enabled; median P50 over 72 hours.
Scoring
Lower seconds is faster.
Maintainer
Artificial Analysis
Reading guide

How to read the scores

Lower median seconds is better. One hop on the default API route with a ~1k prompt; not end-to-end agent or chat-app latency.

Blind spots

What it does not cover

  • End-to-end task time for agent loops — TTFT is one hop, not a full coding session.
  • Non-default providers — AA pins one default route per model for comparability.
Scores

Evidence ledger

186 rows
186186 rows
384.30sbest score
186source-checked
1sources
2026-09-05to 2026-09-01
186

api · api

Distribution

Where the rows land

0255075100

Normalized to this benchmark's axis (0–100). Open the table for raw units.

Timeline

Newest receipts

  1. GPT-6 Astra384.30s ·
  2. Claude Sonnet 5142.07s ·
  3. GPT-5.6 Terra94.25s ·
  4. GPT-5.6 Sol91.64s ·
  5. GPT-5.6 Luna83.33s ·

Scores use this benchmark's own unit and axis, not a universal quality score.

#ModelRelease dateScoreProvenanceTrust
1GPT-6 AstraOpenAIAA FAQ figure for max effort. Includes thinking time before the first answer token.2026-09-03384.30s
2026-09-05
Source-checked
2Claude Sonnet 5Anthropic2026-06-30142.07s
2026-09-01
Source-checked
3GPT-5.6 TerraOpenAI2026-07-0994.25s
2026-09-01
Source-checked
4GPT-5.6 SolOpenAI2026-07-0991.64s
2026-09-01
Source-checked
5GPT-5.6 LunaOpenAI2026-07-0983.33s
2026-09-01
Source-checked
6Claude Fable 5Anthropic2026-06-0954.75s
2026-09-01
Source-checked
7Claude Opus 5 (Adaptive Reasoning, Max Effort)Anthropic2026-07-2436.30s
2026-09-01
Source-checked
8GPT-5.3 Codex (xhigh)OpenAI2026-02-0535.84s
2026-09-01
Source-checked
9Grok 4.6xAI2026-08-1235.04s
2026-09-01
Source-checked
10Claude Opus 5 (Adaptive Reasoning, Xhigh Effort)Anthropic2026-07-2424.72s
2026-09-01
Source-checked
11Gemini 3.1 Pro PreviewGoogle2026-02-1922.68s
2026-09-01
Source-checked
12Grok 4.3 (medium)xAI2026-04-3017.39s
2026-09-01
Source-checked
13Gemini 3.6 Flash (high)Google2026-07-2112.64s
2026-09-01
Source-checked
14Grok 4.5xAI2026-07-0811.48s
2026-09-01
Source-checked
15Claude Opus 5 (Adaptive Reasoning, High Effort)Anthropic2026-07-2410.32s
2026-09-01
Source-checked
16Gemini 3.5 Flash (medium)Google2026-05-198.27s
2026-09-01
Source-checked
17Gemini 3.7 FlashGoogle2026-08-137.00s
2026-09-01
Source-checked
18Gemini 3.5 Flash-LiteGoogle2026-07-216.81s
2026-09-01
Source-checked
19Grok 4.3 (low)xAI2026-04-306.67s
2026-09-01
Source-checked
20Claude Opus 5 (Adaptive Reasoning, Medium Effort)Anthropic2026-07-246.33s
2026-09-01
Source-checked
21OpenAI o3OpenAI2025-04-165.65s
2026-09-01
Source-checked
22MiMo-V2.5Xiaomi2026-04-223.76s
2026-09-01
Source-checked
23MiMo-V2.5-ProXiaomi2026-04-223.74s
2026-09-01
Source-checked
24MiMo-V2.5-Pro (Non-reasoning)Xiaomi2026-04-223.62s
2026-09-01
Source-checked
25Claude Opus 5 (Adaptive Reasoning, Low Effort)Anthropic2026-07-242.66s
2026-09-01
Source-checked
26Kimi K3Moonshot2026-07-162.39s
2026-09-01
Source-checked
27Ring-2.6-1TOther2026-05-082.33s
2026-09-01
Source-checked
28Inkling SmallOther2026-07-302.32s
2026-09-01
Source-checked
29Kimi K3 (low)Moonshot2026-07-162.31s
2026-09-01
Source-checked
30Inkling (xhigh)Other2026-07-152.28s
2026-09-01
Source-checked
31Nemotron 3 Ultra 550B A55B (Reasoning)NVIDIA2026-06-042.23s
2026-09-01
Source-checked
32LongCat-2.0Meituan2026-06-292.13s
2026-09-01
Source-checked
33Ling 3.0 TinyOther2026-08-061.91s
2026-09-01
Source-checked
34Qwen3.8 2.4T A95BAlibaba2026-08-121.76s
2026-09-01
Source-checked
35Step 3.7 FlashStepFun2026-06-011.72s
2026-09-01
Source-checked
36Qwen3.5 397B A17B (Reasoning)Alibaba2026-02-161.71s
2026-09-01
Source-checked
37Hy3Other2026-07-061.69s
2026-09-01
Source-checked
38Qwen3.5 397B A17B (Non-reasoning)Alibaba2026-02-161.66s
2026-09-01
Source-checked
39Agnes 2.5 Pro AlphaOther2026-07-241.63s
2026-09-01
Source-checked
40Ling 3.0 FlashOther2026-08-041.63s
2026-09-01
Source-checked
41GLM-5.2 (Non-reasoning)Zhipu2026-06-161.57s
2026-09-01
Source-checked
42GLM-5.3Zhipu2026-08-141.56s
2026-09-01
Source-checked
43LFM2.5-8B-A1BLiquid2026-06-081.54s
2026-09-01
Source-checked
44Qwen3.6 27B (Reasoning)Alibaba2026-04-221.48s
2026-09-01
Source-checked
45Qwen3.6 27B (Non-reasoning)Alibaba2026-04-221.46s
2026-09-01
Source-checked
46Qwen3.8 MaxAlibaba2026-08-031.46s
2026-09-01
Source-checked
47Gemma 4 12B (Reasoning)Google2026-06-071.45s
2026-09-01
Source-checked
48Qwen 3.7 PlusAlibaba2026-06-021.42s
2026-09-01
Source-checked
49GLM-5.3-FlashZhipu2026-08-261.41s
2026-09-01
Source-checked
50GLM-5.2Zhipu2026-06-131.36s
2026-09-01
Source-checked
51Claude Sonnet 4.6 (Non-reasoning, Low Effort)Anthropic2026-02-171.34s
2026-09-01
Source-checked
52DeepSeek V4 Pro (Non-reasoning)DeepSeek2026-04-241.33s
2026-09-01
Source-checked
53Gemma 4 12B (Non-reasoning)Google2026-06-101.33s
2026-09-01
Source-checked
54Kimi K2.7 CodeMoonshot2026-06-121.30s
2026-09-01
Source-checked
55Qwen3.8-Flash-NextAlibaba2026-08-261.27s
2026-09-01
Source-checked
56Solar Pro 4Other2026-08-061.27s
2026-09-01
Source-checked
57KAT Coder Pro V2Other2026-03-271.26s
2026-09-01
Source-checked
58Qwen3.8 27BAlibaba2026-08-141.21s
2026-09-01
Source-checked
59DeepSeek V4 Pro (Reasoning, High Effort)DeepSeek2026-04-241.16s
2026-09-01
Source-checked
60Agnes 2.5 Pro BetaOther2026-08-261.14s
2026-09-01
Source-checked
61Qwen3.6 35B A3B (Non-reasoning)Alibaba2026-04-161.07s
2026-09-01
Source-checked
62Qwen3.6 35B A3B (Reasoning)Alibaba2026-04-161.06s
2026-09-01
Source-checked
63Qwen3.5 122B A10B (Reasoning)Alibaba2026-02-241.01s
2026-09-01
Source-checked
64DeepSeek V4 Pro 0813 (Reasoning, Max Effort)DeepSeek2026-08-130.98s
2026-09-01
Source-checked
65Gemma 4 31B ITGoogle2026-03-010.97s
2026-09-01
Source-checked
66DeepSeek V4 Flash 0731 (Reasoning, Max Effort)DeepSeek2026-07-310.96s
2026-09-01
Source-checked
67MiniMax M3MiniMax2026-05-310.95s
2026-09-01
Source-checked
68GPT-5.5 Instant (June 2026)OpenAI2026-06-250.93s
2026-09-01
Source-checked
69Nex-N2-ProOther2026-06-020.92s
2026-09-01
Source-checked
70Grok 4.3 (Non-reasoning)xAI2026-04-300.87s
2026-09-01
Source-checked
71Mistral Large 3Mistral2025-12-020.77s
2026-09-01
Source-checked
72Gemini 3.5 Flash (minimal)Google2026-05-190.70s
2026-09-01
Source-checked
73HyperNova 60B 2605Multiverse2026-05-060.69s
2026-09-01
Source-checked
74Mistral Medium 3.5Mistral2026-04-290.64s
2026-09-01
Source-checked
75DeepSeek V4 Flash Vision (Reasoning, Max Effort)DeepSeek2026-08-210.61s
2026-09-01
Source-checked
76Nemotron 3 Nano Omni 30B A3B ReasoningNVIDIA2026-04-290.59s
2026-09-01
Source-checked
77Nemotron 3.5 LightningNVIDIA2026-08-110.55s
2026-09-01
Source-checked
78Llama 4 MaverickMeta2026-04-050.54s
2026-09-01
Source-checked
79Llama 4 ScoutMeta2026-04-050.54s
2026-09-01
Source-checked
80Muse GlimmerMeta2026-08-100.46s
2026-09-01
Source-checked
81Granite 4.2 30BOther2026-08-250.24s
2026-09-01
Source-checked
82Granite 4.2 8BOther2026-08-250.22s
2026-09-01
Source-checked
83North Mini CodeCohere2026-06-090.21s
2026-09-01
Source-checked
84Granite 4.2 3BOther2026-08-250.19s
2026-09-01
Source-checked
85Command ACohere2026-03-150.16s
2026-09-01
Source-checked
86A.X-K2Other2026-08-120.00s
2026-09-01
Source-checked
87Apodex 1.1Other2026-08-300.00s
2026-09-01
Source-checked
88Claude 3 OpusAnthropic2024-03-040.00s
2026-09-01
Source-checked
89Claude 4 Opus (Reasoning)Anthropic2025-05-220.00s
2026-09-01
Source-checked
90Claude 4.1 Opus (Reasoning)Anthropic2025-08-050.00s
2026-09-01
Source-checked
91Claude 4.5 Sonnet (Reasoning)Anthropic2025-09-290.00s
2026-09-01
Source-checked
92Claude Opus 4.5Anthropic2025-11-240.00s
2026-09-01
Source-checked
93Claude Opus 4.5 (Reasoning)Anthropic2025-11-240.00s
2026-09-01
Source-checked
94Claude Opus 4.6Anthropic2026-02-050.00s
2026-09-01
Source-checked
95Claude Opus 4.6 (Adaptive Reasoning, Max Effort)Anthropic2026-02-050.00s
2026-09-01
Source-checked
96Claude Opus 4.7Anthropic2026-04-160.00s
2026-09-01
Source-checked
97Claude Opus 4.7 (Non-reasoning, High Effort)Anthropic2026-04-160.00s
2026-09-01
Source-checked
98Claude Opus 4.8Anthropic2026-05-280.00s
2026-09-01
Source-checked
99Claude Sonnet 4.6Anthropic2026-02-170.00s
2026-09-01
Source-checked
100Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort)Anthropic2026-02-170.00s
2026-09-01
Source-checked
101DeepSeek Coder V2DeepSeek2024-05-150.00s
2026-09-01
Source-checked
102DeepSeek R1DeepSeek2025-01-200.00s
2026-09-01
Source-checked
103DeepSeek V3.2 (Reasoning)DeepSeek2025-12-010.00s
2026-09-01
Source-checked
104DeepSeek V4 Flash (Non-reasoning)DeepSeek2026-04-240.00s
2026-09-01
Source-checked
105DeepSeek V4 Flash (Reasoning, High Effort)DeepSeek2026-04-240.00s
2026-09-01
Source-checked
106DiffusionGemma 26B A4BGoogle2026-06-100.00s
2026-09-01
Source-checked
107EXAONE 4.5 33BOther2026-04-090.00s
2026-09-01
Source-checked
108G9v3-39A5BOther2026-08-030.00s
2026-09-01
Source-checked
109G9v3-3BOther2026-07-230.00s
2026-09-01
Source-checked
110Gemini 1.0 UltraGoogle2024-02-080.00s
2026-09-01
Source-checked
111Gemini 3 Flash PreviewGoogle2025-12-170.00s
2026-09-01
Source-checked
112Gemini 3 Flash Preview (Reasoning)Google2025-12-170.00s
2026-09-01
Source-checked
113Gemini 3 ProGoogle2025-11-180.00s
2026-09-01
Source-checked
114Gemini 3 Pro Preview (low)Google2025-11-180.00s
2026-09-01
Source-checked
115Gemini 3.1 Flash LiteGoogle2026-05-070.00s
2026-09-01
Source-checked
116Gemini 3.5 FlashGoogle2026-05-190.00s
2026-09-01
Source-checked
117GLM 5V Turbo (Reasoning)Zhipu2026-04-010.00s
2026-09-01
Source-checked
118GLM-4.7 (Reasoning)Zhipu2025-12-220.00s
2026-09-01
Source-checked
119GLM-5 (Non-reasoning)Zhipu2026-02-110.00s
2026-09-01
Source-checked
120GLM-5 (Reasoning)Zhipu2026-02-110.00s
2026-09-01
Source-checked
121GLM-5-TurboZhipu2026-03-150.00s
2026-09-01
Source-checked
122GLM-5.1Zhipu2026-04-070.00s
2026-09-01
Source-checked
123GPT 5.4OpenAI2026-03-050.00s
2026-09-01
Source-checked
124GPT 5.5OpenAI2026-04-230.00s
2026-09-01
Source-checked
125GPT-5 (high)OpenAI2025-08-070.00s
2026-09-01
Source-checked
126GPT-5 (low)OpenAI2025-08-070.00s
2026-09-01
Source-checked
127GPT-5 (medium)OpenAI2025-08-070.00s
2026-09-01
Source-checked
128GPT-5 Codex (high)OpenAI2025-09-230.00s
2026-09-01
Source-checked
129GPT-5.1OpenAI2025-11-130.00s
2026-09-01
Source-checked
130GPT-5.1 Codex (high)OpenAI2025-11-130.00s
2026-09-01
Source-checked
131GPT-5.2OpenAI2025-12-110.00s
2026-09-01
Source-checked
132GPT-5.2 (medium)OpenAI2025-12-110.00s
2026-09-01
Source-checked
133GPT-5.2 Codex (xhigh)OpenAI2025-12-110.00s
2026-09-01
Source-checked
134GPT-5.4 (low)OpenAI2026-03-050.00s
2026-09-01
Source-checked
135GPT-5.4 MiniOpenAI2026-03-170.00s
2026-09-01
Source-checked
136GPT-5.4 NanoOpenAI2026-03-170.00s
2026-09-01
Source-checked
137GPT-5.4 ProOpenAI2026-03-050.00s
2026-09-01
Source-checked
138GPT-5.5 (high)OpenAI2026-04-230.00s
2026-09-01
Source-checked
139GPT-5.5 (low)OpenAI2026-04-230.00s
2026-09-01
Source-checked
140GPT-5.5 (medium)OpenAI2026-04-230.00s
2026-09-01
Source-checked
141GPT-5.5 (Non-reasoning)OpenAI2026-04-230.00s
2026-09-01
Source-checked
142GPT-5.5 Instant (May 2026)OpenAI2026-05-050.00s
2026-09-01
Source-checked
143GPT-5.5 ProOpenAI2026-04-230.00s
2026-09-01
Source-checked
144Granite 4.1 30BOther2026-04-290.00s
2026-09-01
Source-checked
145Grok 4xAI2025-07-100.00s
2026-09-01
Source-checked
146Grok 4.20 0309 (Reasoning)xAI2026-03-100.00s
2026-09-01
Source-checked
147Grok 4.20 0309 v2 (Non-reasoning)xAI2026-04-070.00s
2026-09-01
Source-checked
148Grok 4.20 ReasoningxAI2026-03-050.00s
2026-09-01
Source-checked
149Grok 4.3xAI2026-04-300.00s
2026-09-01
Source-checked
150Grok Build 0.1 0616xAI2026-06-160.00s
2026-09-01
Source-checked
151Hy3-preview (Non-reasoning)Other2026-04-230.00s
2026-09-01
Source-checked
152Hy3-preview (Reasoning)Other2026-04-230.00s
2026-09-01
Source-checked
153JT-35B-FlashOther2026-05-140.00s
2026-09-01
Source-checked
154JT-4.1 Flash 236B A21BOther2026-07-090.00s
2026-09-01
Source-checked
155JT-MINIOther2026-04-150.00s
2026-09-01
Source-checked
156K-EXAONE 2.0 0803Other2026-08-120.00s
2026-09-01
Source-checked
157Kimi K2 ThinkingMoonshot2025-11-060.00s
2026-09-01
Source-checked
158Kimi K2.5Moonshot2026-01-270.00s
2026-09-01
Source-checked
159Kimi K2.6Moonshot2026-04-200.00s
2026-09-01
Source-checked
160Kimi K2.6 (Non-reasoning)Other2026-04-200.00s
2026-09-01
Source-checked
161Ling 2.6 FlashOther2026-04-210.00s
2026-09-01
Source-checked
162Ling-2.6-1TOther2026-04-230.00s
2026-09-01
Source-checked
163MiMo-V2-Flash (Feb 2026)Xiaomi2025-12-160.00s
2026-09-01
Source-checked
164MiMo-V2-Flash (Reasoning)Xiaomi2025-12-160.00s
2026-09-01
Source-checked
165MiMo-V2-OmniXiaomi2026-03-190.00s
2026-09-01
Source-checked
166MiMo-V2-Omni-0327Xiaomi2026-03-270.00s
2026-09-01
Source-checked
167MiMo-V2-ProXiaomi2026-03-180.00s
2026-09-01
Source-checked
168MiniCPM-V 4.6 1.3BOpenBMB2026-05-110.00s
2026-09-01
Source-checked
169MiniCPM5-1B (Non-reasoning)OpenBMB2026-05-250.00s
2026-09-01
Source-checked
170MiniCPM5-1B (Reasoning)OpenBMB2026-06-040.00s
2026-09-01
Source-checked
171MiniMax M2.5MiniMax2026-04-010.00s
2026-09-01
Source-checked
172MiniMax M2.7MiniMax2026-04-150.00s
2026-09-01
Source-checked
173MiniMax-M2.1MiniMax2025-12-230.00s
2026-09-01
Source-checked
174Motif 3Other2026-08-120.00s
2026-09-01
Source-checked
175Motif 3 (Beta)Other2026-07-210.00s
2026-09-01
Source-checked
176Muse SparkOther2026-01-010.00s
2026-09-01
Source-checked
177Muse Spark 1.1Meta2026-07-090.00s
2026-09-01
Source-checked
178Muse Spark 1.2Meta2026-08-050.00s
2026-09-01
Source-checked
179o3-proOpenAI2025-06-100.00s
2026-09-01
Source-checked
180OpenAI o1OpenAI2024-09-120.00s
2026-09-01
Source-checked
181Qwen3 Max ThinkingAlibaba2026-01-260.00s
2026-09-01
Source-checked
182Qwen3.5 27B (Reasoning)Alibaba2026-02-240.00s
2026-09-01
Source-checked
183Qwen3.6 Max PreviewAlibaba2026-04-200.00s
2026-09-01
Source-checked
184Qwen3.6 PlusAlibaba2026-04-020.00s
2026-09-01
Source-checked
185Qwen3.7 MaxAlibaba2026-05-190.00s
2026-09-01
Source-checked
186Solar Open2 250BOther2026-08-120.00s
2026-09-01
Source-checked
Method

What it covers

This benchmark sits in the performance family. It uses Single-shot API request with streaming enabled; median P50 over 72 hours. The score should travel with its task format, scoring method, source date, and benchmark version.

Score ceiling

Where it breaks down

No ceiling note is recorded yet. Treat clustering near the top as a warning that the benchmark may no longer separate frontier models.

Tools that report it

Receipts

Sources and further reading

← All benchmarksHow we score

Questions about this benchmark

Every answer below is assembled from the dated fields on this page. Nothing is written separately for search.

What does AA time to first token measure?

Median seconds until the first token returns on the model's default API provider with a ~1k-token input prompt.

What does a high AA time to first token score not prove?

A strong AA time to first token result says nothing about:

  • End-to-end task time for agent loops — TTFT is one hop, not a full coding session.
  • Non-default providers — AA pins one default route per model for comparability.

How is AA time to first token scored?

Lower seconds is faster. Task format: Single-shot API request with streaming enabled; median P50 over 72 hours.

Is AA time to first token saturated?

AA time to first token is currently marked Active in the atlas.

Can AA time to first token results be contaminated by training data?

Contamination risk for AA time to first token is graded Unknown contamination. Treat every row on this page as a public claim with a source and a date, not as a controlled experiment.

The Pack · Editorial newsletter

New tools in your inbox. Free.

One short email when a tool ships or changes status. No tracking, no third-party analytics. Unsubscribe in one click.