Sign inTest your prompt

Benchmarks / BulletBench

Measured by Spring Prompt

BulletBench

When every second of thinking time comes off the clock, which models are fast enough to still make good decisions?

Last updated 13 Nov 2026

Results dated
1 Oct 2026 to 13 Nov 2026
Results
70 configurations of 58 models
Unit
% of moves
Licence
Spring Prompt original
Sample
8 to 12 games per configuration

BulletBench in brief

What it measures. BulletBench is a Spring Prompt benchmark. When thinking time comes off the clock, which models are quick enough to still make good decisions? It scores: Decision quality under real time pressure; Response latency at the provider's default settings; Which reasoning settings are fast enough to use.

Latest results. , 70 configurations ranked.

Top 3 on Invalid moves, Bullet 60s (% of moves, lower is better):

  1. Claude Fable 5.2 (low reasoning) (Anthropic): 0.0%
  2. Claude Fable 5.2 (Anthropic): 0.0%
  3. Claude Haiku 5.5 (no reasoning) (Anthropic): 0.0%

The first 61 are too close to tell apart, so read the order among them as a tie.

How to cite. Spring Prompt, "BulletBench", results dated 13 November 2026. https://springprompt.com/benchmarks/bulletbench

Invalid moves:

9 results above zero · % of moves, lower is better. Choose a model to highlight it.Clear highlight

  1. 62 Gemini 3.6 Flash (minimal reasoning)Google 0.3%
  2. 62 Claude Haiku 4.5Anthropic 0.3%
  3. 64 Seed-2.0-Mini (minimal reasoning)ByteDance 0.4%
  4. 64 Mistral Medium 3.5 (no reasoning)Mistral AI 0.4%
  5. 66 Gemini 3.5 Flash (minimal reasoning)Google 0.5%
  6. 67 Mistral Small 4 (no reasoning)Mistral AI 0.6%
  7. 68 Mistral Medium 3.5Mistral AI 0.8%
  8. 69 GPT-5.4 nano (no reasoning)OpenAI 3.1%
  9. 70 Inkling Small (no reasoning)Thinking Machines 7.2%

61 models had none: Claude Fable 5.2, Claude Fable 5.2 (low reasoning), Claude Haiku 5.5 (no reasoning), Clef, Clef Flash, Clef Omni, Decider Flash v1.2, Decider V1 27B, Decider v1.2, DeepSeek-V4.1-Flash-1017 (low reasoning), Drex Max, Drex Max Finance, Drex v1.5, Drex v2.0, Elephant 1, GLM-5.3, GLM-5.3 (low reasoning), GLM-5.3 FlashX Legal (low reasoning), GPT-5.4 mini (no reasoning), GPT-6 Astra, GPT-6 Astra (low reasoning), GPT-6 Luna (no reasoning), GPT-6 Luna Decisions, GPT-6 Sol (no reasoning), GPT-6.1 Sol, GPT-6.1 Sol (low reasoning), GPT-6.1 Sol Ultrafast, GPT-6.1 Sol Ultrafast (low reasoning), GPT-6.1 Terra (low reasoning), Gemini 3.1 Flash-Lite (minimal reasoning), Gemini 3.1 Pro Preview (low reasoning), Gemini 3.5 Flash-Lite, Gemini 3.8 Flash, Gemini 3.8 Flash (low reasoning), Gemini 4 Argon, Gemini 4 Carbon (low reasoning), Gemini 4 Hydrogen, Gemini 4 Hydrogen (low reasoning), Gemini Decision 1, Jev 1.13, Jev 1.14, Joey 1, Kimi K3 (low reasoning), Kimi K3.1 Pro (low reasoning), Mammoth 1, Mercury 3 (low reasoning), Mercury 3 (no reasoning), Mercury Decide, Microsoft Decision 1, MiniMax M3.1 Flash (low reasoning), Muse Ember, Muse Ember (low reasoning), Muse Ember (minimal reasoning), Muse Spark 1.3 (minimal reasoning), Muse Wonder (minimal reasoning), Parallel Decisions Fast, Reflex, Span-02, d1, d1 Ultra, d1.1.

Lightning: strength against thinking time

Each configuration's median time per move at 10 seconds plus 1 a move. Right of the line, a model thinks longer than the increment it earns and the clock drains.

Named: the six bestOther models (hover for names)

-200300800 0.1 s1 s10 s Median time per move (log scale) Ladder Elo 1 s increment Gemini 3.5 Flash-Lite Gemini 3.1 Flash-Lite (minimal reasoning) Gemini Decision 1 Gemini 3.5 Flash (minimal reasoning) Clef Omni Reflex

Full results

BulletBench Bullet 60s: invalid moves, % of moves, lower is better
#ModelInvalid moves, Bullet 60s
% of moves, lower is better		Lightning 10+1
ladder Elo		Bullet 60s
ladder Elo		Blitz 3+2
ladder Elo		Move time
seconds		Lost on time
% of games		Cost per game
US dollars		Slowest 10% of moves
seconds, 90th percentile		Games
at this clock
1≈ Claude Fable 5.2 (low reasoning)Anthropic
0.0%
–491–3.0 s62.5%$0.0800 4.3 s8
1≈ Claude Fable 5.2Anthropic
0.0%
–4911,1003.4 s62.5%$0.0948 5.3 s8
1≈ Claude Haiku 5.5 (no reasoning)Anthropic
0.0%
–310–1.7 s37.5%$0.0019 2.5 s8
1≈ Elephant 1Decision modelBaseten
0.0%
614571–0.5 s0.0%$0.0008 0.6 s12
1≈ Joey 1Decision modelBaseten
0.0%
529529–0.5 s0.0%$0.0003 0.5 s12
1≈ Mammoth 1Decision modelBaseten
0.0%
571571–0.6 s0.0%$0.0017 0.6 s12
1≈ Clef FlashDecision modelCloudflare
0.0%
529529–0.5 s0.0%$0.0025 0.7 s12
1≈ Clef OmniDecision modelCloudflare
0.0%
616571–0.5 s0.0%$0.0049 0.6 s12
1≈ ClefDecision modelCloudflare
0.0%
529529–0.7 s0.0%$0.0056 1.0 s12
1≈ DeepSeek-V4.1-Flash-1017 (low reasoning)DeepSeek
0.0%
–147–3.3 s87.5%$0.0047 ––
1≈ Gemini 3.1 Flash-Lite (minimal reasoning)Google
0.0%
780912–0.8 s8.3%$0.0068 ––
1≈ Gemini 3.1 Pro Preview (low reasoning)Google
0.0%
–529–3.1 s58.3%$0.0446 4.4 s12
1≈ Gemini 3.5 Flash-LiteGoogle
0.0%
8927807070.9 s0.0%$0.0058 ––
1≈ Gemini 3.8 Flash (low reasoning)Google
0.0%
1971,068–1.8 s50.0%$0.0165 2.2 s12
1≈ Gemini 3.8 FlashGoogle
0.0%
–5291,1262.8 s58.3%$0.0154 4.8 s12
1≈ Gemini 4 ArgonGoogle
0.0%
–1471,1472.9 s87.5%$0.0036 9.8 s8
1≈ Gemini 4 Carbon (low reasoning)Google
0.0%
–3588483.2 s75.0%$0.0167 5.9 s8
1≈ Gemini 4 Hydrogen (low reasoning)Google
0.0%
–557–2.6 s62.5%$0.0439 ––
1≈ Gemini 4 HydrogenGoogle
0.0%
–5571,0523.4 s62.5%$0.0451 ––
1≈ Gemini Decision 1Decision modelGoogle
0.0%
755571–0.7 s0.0%$0.0023 0.8 s12
1≈ Mercury 3 (low reasoning)Inception
0.0%
–147–4.2 s87.5%$0.0032 16.1 s8
1≈ Mercury 3 (no reasoning)Inception
0.0%
555491–0.4 s0.0%$0.0020 ––
1≈ Mercury DecideDecision modelInception
0.0%
614572–0.4 s0.0%$0 0.5 s12
1≈ d1.1Decision modelLiquid AI
0.0%
572614–0.6 s0.0%$0.0003 0.8 s12
1≈ d1 UltraDecision modelLiquid AI
0.0%
529572–0.6 s0.0%$0.0003 1.0 s12
1≈ d1Decision modelLiquid AI
0.0%
572584–0.6 s0.0%$0.0007 0.7 s12
1≈ Muse Ember (low reasoning)Meta
0.0%
–147–2.8 s75.0%$0.0106 8.4 s8
1≈ Muse Ember (minimal reasoning)Meta
0.0%
–310–1.8 s62.5%$0.0038 3.2 s8
1≈ Muse EmberMeta
0.0%
–147–3.4 s87.5%$0.0108 22.2 s8
1≈ Muse Spark 1.3 (minimal reasoning)Meta
0.0%
–33–4.2 s91.7%$0.0142 11.4 s12
1≈ Muse Wonder (minimal reasoning)Meta
0.0%
–409–1.9 s58.3%$0.0663 4.1 s12
1≈ Microsoft Decision 1Decision modelMicrosoft
0.0%
453490–0.6 s0.0%$0.0009 0.7 s12
1≈ MiniMax M3.1 Flash (low reasoning)MiniMax
0.0%
–115–3.0 s75.0%$0.0016 9.3 s8
1≈ Kimi K3.1 Pro (low reasoning)Moonshot AI
0.0%
–147–5.0 s87.5%$0.0175 18.1 s8
1≈ Kimi K3 (low reasoning)Moonshot AI
0.0%
–33–2.6 s91.7%$0.0540 11.3 s12
1≈ ReflexDecision modelMorph
0.0%
616614–0.6 s0.0%$0.0006 0.7 s12
1≈ Drex Max FinanceDecision modelNace AI
0.0%
614571–0.7 s0.0%$0.0010 1.2 s12
1≈ Drex MaxDecision modelNace AI
0.0%
409572–1.2 s33.3%$0.0011 2.0 s12
1≈ Drex v1.5Decision modelNace AI
0.0%
489449–0.6 s8.3%$0.0007 0.9 s12
1≈ Drex v2.0Decision modelNace AI
0.0%
571571–0.6 s0.0%$0.0009 1.0 s12
1≈ GPT-5.4 mini (no reasoning)OpenAI
0.0%
571197–0.9 s50.0%$0.0205 ––
1≈ GPT-6.1 Sol Ultrafast (low reasoning)OpenAI
0.0%
584928–1.4 s25.0%$0.34 ––
1≈ GPT-6.1 Sol (low reasoning)OpenAI
0.0%
–323–2.2 s75.0%$0.0243 ––
1≈ GPT-6.1 Sol UltrafastOpenAI
0.0%
1975409271.7 s41.7%$0.36 3.0 s12
1≈ GPT-6.1 SolOpenAI
0.0%
–336453.0 s91.7%$0.0210 8.7 s12
1≈ GPT-6.1 Terra (low reasoning)OpenAI
0.0%
491358–2.8 s75.0%$0.0172 5.1 s8
1≈ GPT-6 Astra (low reasoning)OpenAI
0.0%
–323–2.3 s75.0%$0.10 5.7 s12
1≈ GPT-6 AstraOpenAI
0.0%
–1978393.0 s83.3%$0.0943 6.1 s12
1≈ GPT-6 Luna DecisionsDecision modelOpenAI
0.0%
529489–0.5 s0.0%$0.0026 0.7 s12
1≈ GPT-6 Luna (no reasoning)OpenAI
0.0%
–249–1.5 s50.0%$0.0016 2.3 s12
1≈ GPT-6 Sol (no reasoning)OpenAI
0.0%
–445–1.6 s50.0%$0.0297 2.1 s12
1≈ Parallel Decisions FastDecision modelParallel
0.0%
529571–0.8 s0.0%$0 1.0 s12
1≈ Decider Flash v1.2Decision modelPerplexity
0.0%
489489–0.5 s0.0%$0.0003 0.6 s12
1≈ Decider V1 27BDecision modelPerplexity
0.0%
420385–0.6 s8.3%$0.0010 0.8 s12
1≈ Decider v1.2Decision modelPerplexity
0.0%
572571–0.6 s0.0%$0.0003 0.9 s12
1≈ Span-02Decision modelRespan
0.0%
614571–0.9 s0.0%$0.0002 1.2 s12
1≈ Jev 1.13Decision modelTypeSafe
0.0%
571572–0.5 s0.0%$0.0013 0.6 s12
1≈ Jev 1.14Decision modelTypeSafe
0.0%
529529–0.5 s0.0%$0.0014 0.5 s12
1≈ GLM-5.3 FlashX Legal (low reasoning)Z.ai
0.0%
310358–1.4 s37.5%$0.0061 3.0 s8
1≈ GLM-5.3 (low reasoning)Z.ai
0.0%
–33–1.6 s83.3%$0.0220 ––
1≈ GLM-5.3Z.ai
0.0%
–33–3.6 s91.7%$0.0156 ––
62 Gemini 3.6 Flash (minimal reasoning)Google
0.3%
33628–1.6 s41.7%$0.0113 1.9 s12
62 Claude Haiku 4.5Anthropic
0.3%
334454911.2 s25.0%$0.0163 1.7 s12
64 Seed-2.0-Mini (minimal reasoning)ByteDance
0.4%
571453–0.3 s0.0%$0.0030 0.4 s12
64 Mistral Medium 3.5 (no reasoning)Mistral AI
0.4%
529572–0.5 s0.0%$0.0354 0.7 s12
66 Gemini 3.5 Flash (minimal reasoning)Google
0.5%
727810–1.2 s25.0%$0.0289 1.4 s12
67 Mistral Small 4 (no reasoning)Mistral AI
0.6%
571367–0.6 s25.0%$0.0039 1.1 s12
68 Mistral Medium 3.5Mistral AI
0.8%
5293855550.5 s8.3%$0.0420 1.0 s12
69 GPT-5.4 nano (no reasoning)OpenAI
3.1%
19733–1.1 s66.7%$0.0050 ––
70 Inkling Small (no reasoning)Thinking Machines
7.2%
571529–0.6 s0.0%$0.0075 1.5 s12

Swipe the table sideways for more columns.

Source: Spring Prompt, springprompt.com/benchmarks/bulletbench. Test your own prompt on these models →

Ranks follow the score as shown, so equal numbers share a rank. Each model runs at its provider's default reasoning setting and, where the provider offers a faster one, again at its fastest setting, listed separately. Blitz 3+2 was run at default settings only. Equal ratings are not a glitch: the ladder's ratings come in steps (see Honest caveats below).

Progress over time

Invalid moves, plotted at each model's release date, lower is better. Each model counts once, at its best setting. Scores are today's, not launch day's; 27 models without a known release date are left off. Hover or tap a mark for its model.

Best score so far

Every model at its release date. The line moves each time a new release beats every earlier one, and each new best is named where there is room.

-2%0%2%4%6%8% 2026AprJulOct Claude Haiku 4.5Gemini 3.1 Pro Preview

Best score by lab

Each lab's best so far, stepping whenever one of its releases beats its own earlier models. The 6 labs with the best scores today; 5 others not drawn.

AnthropicGoogleOpenAIMoonshot AIZ.aiMeta

-0.1%0%0.1%0.2%0.3%0.4% 2026AprJulOct AnthropicGoogleOpenAIMoonshot AIZ.aiMeta

Too slow for the clock

These configurations lost their first four games on time, so they were stopped there and are not rated for that clock. Each shows its median time per move and the time that 1 move in 10 took or exceeded: a median under the increment still loses on time when the slow tail is long. Being too slow is the result, not a fault.

Bullet 60s

  • Claude Fable 5.1 median 4.2 s, 1 in 10 moves 6.3 s+
  • Claude Fable 5.1 (low reasoning) median 4.3 s, 1 in 10 moves 9.3 s+
  • Claude Haiku 5.5 median 4.6 s, 1 in 10 moves 8.1 s+
  • Claude Haiku 5.5 (low reasoning) median 3.6 s, 1 in 10 moves 9.5 s+
  • Claude Opus 5.5 median 3.1 s, 1 in 10 moves 5.8 s+
  • Claude Opus 5.5 (low reasoning) median 3.1 s, 1 in 10 moves 5.3 s+
  • Claude Sonnet 5.5 median 3.0 s, 1 in 10 moves 9.4 s+
  • Claude Sonnet 5.5 (low reasoning) median 2.6 s, 1 in 10 moves 5.0 s+
  • DeepSeek V4 Pro 0423 median 5.0 s, 1 in 10 moves 27.8 s+
  • DeepSeek V4.1 Flash (low reasoning) median 1.9 s, 1 in 10 moves 10.6 s+
  • DeepSeek V4.1 Flash 1017 median 2.4 s, 1 in 10 moves 20.0 s+
  • Gemini 3.1 Pro Preview median 4.9 s, 1 in 10 moves 8.4 s+
  • Gemini 4 Argon (low reasoning) median 5.0 s, 1 in 10 moves 5.0 s+
  • Gemini 4 Carbon median 5.1 s, 1 in 10 moves 8.7 s+
  • Mercury 3 median 4.2 s, 1 in 10 moves 25.6 s+
  • Kev 4B median 1.9 s, 1 in 10 moves 15.4 s+
  • Muse Spark 1.3 median 7.0 s, 1 in 10 moves 28.1 s+
  • Muse Wonder median 3.6 s, 1 in 10 moves 8.1 s+
  • MiniMax M3.1 Flash median 4.1 s, 1 in 10 moves 17.9 s+
  • minimax/minimax-m3.1-flash:free median 3.9 s, 1 in 10 moves 20.8 s+
  • minimax/minimax-m3.1-flash:free (low reasoning) median 3.0 s, 1 in 10 moves 8.4 s+
  • Mistral Large 4 median 0.8 s, 1 in 10 moves 2.9 s+
  • Kimi K3 median 3.2 s, 1 in 10 moves 33.8 s+
  • Kimi K3.1 Pro median 6.1 s, 1 in 10 moves 16.1 s+
  • GPT-6 Luna median 4.6 s, 1 in 10 moves 10.8 s+
  • GPT-6 Sol median 4.2 s, 1 in 10 moves 9.4 s+
  • GPT-6.1 Terra median 3.4 s, 1 in 10 moves 7.3 s+
  • Particle-1 median 4.3 s, 1 in 10 moves 12.7 s+
  • Particle-1 (low reasoning) median 4.4 s, 1 in 10 moves 12.8 s+
  • Qwen3.8 Flash median 3.8 s, 1 in 10 moves 16.9 s+
  • Qwen3.8 Max (0902) median 4.9 s, 1 in 10 moves 22.8 s+
  • Qwen3.8 Max (0902) (minimal reasoning) median 4.7 s, 1 in 10 moves 10.3 s+
  • Grok 4.7 median 4.7 s, 1 in 10 moves 17.7 s+
  • Grok 4.7 (low reasoning) median 5.4 s, 1 in 10 moves 15.5 s+
  • MiMo-V2.6-Flash median 4.0 s, 1 in 10 moves 11.7 s+
  • GLM 4.7 Flash median 9.7 s, 1 in 10 moves 23.1 s+
  • GLM 5.3 Flash (low reasoning) median 2.2 s, 1 in 10 moves 5.3 s+
  • GLM 5.3 FlashX Legal median 2.1 s, 1 in 10 moves 11.0 s+

Lightning 10+1

  • Claude Fable 5.1 median 4.4 s, 1 in 10 moves 6.7 s+
  • Claude Fable 5.1 (low reasoning) median 4.1 s, 1 in 10 moves 6.7 s+
  • Claude Fable 5.2 median 2.2 s, 1 in 10 moves 4.1 s+
  • Claude Fable 5.2 (low reasoning) median 3.7 s, 1 in 10 moves 4.4 s+
  • Claude Haiku 5.5 median 1.9 s, 1 in 10 moves 4.8 s+
  • Claude Haiku 5.5 (low reasoning) median 1.9 s, 1 in 10 moves 2.9 s+
  • Claude Haiku 5.5 (no reasoning) median 1.8 s, 1 in 10 moves 2.2 s+
  • Claude Opus 5.5 median 3.0 s, 1 in 10 moves 3.5 s+
  • Claude Opus 5.5 (low reasoning) median 2.8 s, 1 in 10 moves 3.3 s+
  • Claude Sonnet 5.5 median 2.1 s, 1 in 10 moves 2.7 s+
  • DeepSeek V4 Pro 0423 median 4.6 s, 1 in 10 moves 6.6 s+
  • DeepSeek V4.1 Flash (low reasoning) median 0.8 s, 1 in 10 moves 1.7 s+
  • DeepSeek V4.1 Flash 1017 median 1.4 s, 1 in 10 moves 4.9 s+
  • DeepSeek V4.1 Flash 1017 (low reasoning) median 1.2 s, 1 in 10 moves 5.3 s+
  • Gemini 3.1 Pro Preview median 2.7 s, 1 in 10 moves 3.4 s+
  • Gemini 3.1 Pro Preview (low reasoning) median 2.5 s, 1 in 10 moves 3.0 s+
  • Gemini 3.8 Flash median 2.3 s, 1 in 10 moves 2.9 s+
  • Gemini 4 Argon median 3.5 s, 1 in 10 moves 5.2 s+
  • Gemini 4 Argon (low reasoning) median 2.6 s, 1 in 10 moves 6.6 s+
  • Gemini 4 Carbon median 3.6 s, 1 in 10 moves 4.1 s+
  • Gemini 4 Carbon (low reasoning) median 2.6 s, 1 in 10 moves 4.4 s+
  • Gemini 4 Hydrogen median 1.8 s
  • Gemini 4 Hydrogen (low reasoning) median 1.9 s
  • Mercury 3 median 0.9 s, 1 in 10 moves 2.7 s+
  • Mercury 3 (low reasoning) median 1.2 s, 1 in 10 moves 4.7 s+
  • Muse Ember median 1.7 s, 1 in 10 moves 5.0 s+
  • Muse Ember (low reasoning) median 1.7 s, 1 in 10 moves 2.9 s+
  • Muse Ember (minimal reasoning) median 1.4 s, 1 in 10 moves 2.6 s+
  • Muse Spark 1.3 median 4.6 s, 1 in 10 moves 9.5 s+
  • Muse Spark 1.3 (minimal reasoning) median 2.2 s, 1 in 10 moves 4.9 s+
  • Muse Wonder median 2.9 s, 1 in 10 moves 3.9 s+
  • Muse Wonder (minimal reasoning) median 1.6 s, 1 in 10 moves 2.8 s+
  • MiniMax M3.1 Flash median 2.8 s, 1 in 10 moves 4.5 s+
  • MiniMax M3.1 Flash (low reasoning) median 2.2 s, 1 in 10 moves 3.4 s+
  • minimax/minimax-m3.1-flash:free median 1.9 s, 1 in 10 moves 3.8 s+
  • minimax/minimax-m3.1-flash:free (low reasoning) median 1.8 s, 1 in 10 moves 3.9 s+
  • Kimi K3 median 1.4 s, 1 in 10 moves 6.5 s+
  • Kimi K3 (low reasoning) median 1.7 s, 1 in 10 moves 4.2 s+
  • Kimi K3.1 Pro median 4.5 s, 1 in 10 moves 5.8 s+
  • Kimi K3.1 Pro (low reasoning) median 3.6 s, 1 in 10 moves 4.6 s+
  • GPT-6 Astra median 2.0 s, 1 in 10 moves 3.4 s+
  • GPT-6 Astra (low reasoning) median 2.1 s, 1 in 10 moves 3.1 s+
  • GPT-6 Luna median 2.9 s, 1 in 10 moves 4.2 s+
  • GPT-6 Luna (no reasoning) median 2.1 s, 1 in 10 moves 3.1 s+
  • GPT-6 Sol median 1.6 s, 1 in 10 moves 3.7 s+
  • GPT-6 Sol (no reasoning) median 1.6 s, 1 in 10 moves 2.0 s+
  • GPT-6.1 Sol median 2.0 s, 1 in 10 moves 3.4 s+
  • GPT-6.1 Sol (low reasoning) median 1.6 s, 1 in 10 moves 3.2 s+
  • GPT-6.1 Terra median 1.3 s, 1 in 10 moves 3.4 s+
  • Particle-1 median 2.2 s, 1 in 10 moves 3.5 s+
  • Particle-1 (low reasoning) median 1.3 s, 1 in 10 moves 3.6 s+
  • Qwen3.8 Flash median 2.5 s, 1 in 10 moves 3.8 s+
  • Qwen3.8 Max (0902) median 3.1 s, 1 in 10 moves 4.6 s+
  • Qwen3.8 Max (0902) (minimal reasoning) median 2.9 s, 1 in 10 moves 4.9 s+
  • Grok 4.7 median 2.5 s, 1 in 10 moves 5.5 s+
  • Grok 4.7 (low reasoning) median 2.3 s, 1 in 10 moves 4.3 s+
  • MiMo-V2.6-Flash median 3.1 s, 1 in 10 moves 4.6 s+
  • GLM 4.7 Flash median 4.4 s, 1 in 10 moves 6.0 s+
  • GLM 5.3 median 1.8 s, 1 in 10 moves 6.2 s+
  • GLM 5.3 (low reasoning) median 1.2 s, 1 in 10 moves 4.0 s+
  • GLM 5.3 Flash (low reasoning) median 0.9 s, 1 in 10 moves 2.6 s+
  • GLM 5.3 FlashX Legal median 1.3 s, 1 in 10 moves 3.1 s+

Blitz 3+2

  • DeepSeek V4 Pro 0423 median 6.4 s, 1 in 10 moves 29.7 s+
  • DeepSeek V4.1 Flash 1017 median 4.7 s, 1 in 10 moves 62.2 s+
  • Mercury 3 median 9.0 s, 1 in 10 moves 72.7 s+
  • Muse Ember median 6.1 s, 1 in 10 moves 48.0 s+
  • Muse Spark 1.3 median 19.4 s, 1 in 10 moves 60.0 s+
  • MiniMax M3.1 Flash median 9.1 s, 1 in 10 moves 46.9 s+
  • minimax/minimax-m3.1-flash:free median 5.3 s, 1 in 10 moves 76.0 s+
  • Mistral Large 4 (high reasoning) median 26.1 s, 1 in 10 moves 69.5 s+
  • Kimi K3 median 6.8 s, 1 in 10 moves 32.1 s+
  • Kimi K3.1 Pro median 10.8 s, 1 in 10 moves 47.7 s+
  • GPT-6 Sol median 8.2 s, 1 in 10 moves 22.9 s+
  • Particle-1 median 5.0 s, 1 in 10 moves 39.3 s+
  • Qwen3.8 Max (0902) median 7.4 s, 1 in 10 moves 88.9 s+
  • Grok 4.7 median 9.9 s, 1 in 10 moves 32.9 s+
  • GLM 5.3 FlashX Legal median 3.0 s, 1 in 10 moves 33.7 s+

Decision models

Decision models answer a typed question instead of writing text. Each move is one question whose options are the legal moves, so they cannot play an illegal move, and they are not told their clock. They play the same clock, ladder and engine and are in the table above; here they are ranked among themselves.

BulletBench Bullet 60s: invalid moves, % of moves, lower is better
#ModelInvalid moves, Bullet 60s
% of moves, lower is better		Lightning 10+1
ladder Elo		Bullet 60s
ladder Elo		Move time
seconds		Lost on time
% of games		Cost per game
US dollars		Slowest 10% of moves
seconds, 90th percentile		Games
at this clock
1≈ Elephant 1Baseten
0.0%
6145710.5 s0.0%$0.0008 0.6 s12
1≈ Joey 1Baseten
0.0%
5295290.5 s0.0%$0.0003 0.5 s12
1≈ Mammoth 1Baseten
0.0%
5715710.6 s0.0%$0.0017 0.6 s12
1≈ Clef FlashCloudflare
0.0%
5295290.5 s0.0%$0.0025 0.7 s12
1≈ Clef OmniCloudflare
0.0%
6165710.5 s0.0%$0.0049 0.6 s12
1≈ ClefCloudflare
0.0%
5295290.7 s0.0%$0.0056 1.0 s12
1≈ Gemini Decision 1Google
0.0%
7555710.7 s0.0%$0.0023 0.8 s12
1≈ Mercury DecideInception
0.0%
6145720.4 s0.0%$0 0.5 s12
1≈ d1.1Liquid AI
0.0%
5726140.6 s0.0%$0.0003 0.8 s12
1≈ d1 UltraLiquid AI
0.0%
5295720.6 s0.0%$0.0003 1.0 s12
1≈ d1Liquid AI
0.0%
5725840.6 s0.0%$0.0007 0.7 s12
1≈ Microsoft Decision 1Microsoft
0.0%
4534900.6 s0.0%$0.0009 0.7 s12
1≈ ReflexMorph
0.0%
6166140.6 s0.0%$0.0006 0.7 s12
1≈ Drex Max FinanceNace AI
0.0%
6145710.7 s0.0%$0.0010 1.2 s12
1≈ Drex MaxNace AI
0.0%
4095721.2 s33.3%$0.0011 2.0 s12
1≈ Drex v1.5Nace AI
0.0%
4894490.6 s8.3%$0.0007 0.9 s12
1≈ Drex v2.0Nace AI
0.0%
5715710.6 s0.0%$0.0009 1.0 s12
1≈ GPT-6 Luna DecisionsOpenAI
0.0%
5294890.5 s0.0%$0.0026 0.7 s12
1≈ Parallel Decisions FastParallel
0.0%
5295710.8 s0.0%$0 1.0 s12
1≈ Decider Flash v1.2Perplexity
0.0%
4894890.5 s0.0%$0.0003 0.6 s12
1≈ Decider V1 27BPerplexity
0.0%
4203850.6 s8.3%$0.0010 0.8 s12
1≈ Decider v1.2Perplexity
0.0%
5725710.6 s0.0%$0.0003 0.9 s12
1≈ Span-02Respan
0.0%
6145710.9 s0.0%$0.0002 1.2 s12
1≈ Jev 1.13TypeSafe
0.0%
5715720.5 s0.0%$0.0013 0.6 s12
1≈ Jev 1.14TypeSafe
0.0%
5295290.5 s0.0%$0.0014 0.5 s12

Swipe the table sideways for more columns.

Source: Spring Prompt, springprompt.com/benchmarks/bulletbench. Test your own prompt on these models →

Too slow for the clock. These lost their first four games on time and are not rated for that clock.

Bullet 60s

  • Kev 4B median 1.9 s, 1 in 10 moves 15.4 s+

Real games: the clock at Lightning 10+1

Each model's clock after every move of one game. A fast model earns back more than it spends; a slow one runs out within a few moves, whatever the position.

0 s5 s10 s15 s 0481216202428 Move Clock left Won by checkmate after 28 moves Gemini 3.5 Flash Lite Lost on time after 3 moves Claude Opus 5.5
Gemini 3.5 Flash LiteWon by checkmate after 28 moves
  1. e40.6 s10.4 s left
  2. Nf30.9 s10.4 s left
  3. Bc40.6 s10.9 s left
  4. O-O0.9 s10.9 s left
  5. c30.6 s11.4 s left
  6. d40.9 s11.4 s left
  7. Nxe51.0 s11.4 s left
  8. Nxc60.6 s11.8 s left
  9. Qxg40.9 s11.9 s left
  10. Qxc8+0.9 s12.0 s left
  11. … 18 more moves
Claude Opus 5.5Lost on time after 3 moves
  1. e43.5 s7.5 s left
  2. Nf33.5 s5.0 s left
  3. d43.3 s2.6 s left

More from the results

Bullet: strength against thinking time

The same at 60 seconds for the whole game, no increment. A 40-move game leaves 1.5 seconds a move.

Named: the six bestOther models (hover for names)

-50005001,0001,500 0.1 s1 s10 s Median time per move (log scale) Ladder Elo 1.5 s a move Gemini 3.8 Flash (low reasoning) GPT-6.1 Sol Ultrafast (low reasoning) Gemini 3.1 Flash-Lite (minimal reasoning) Gemini 3.5 Flash (minimal reasoning) Gemini 3.5 Flash-Lite Gemini 3.6 Flash (minimal reasoning)

How games are lost

Share of games lost on time, and moves with no legal reply, by clock.

ModelOn time, LightningOn time, BulletInvalid moves, LightningInvalid moves, Bullet
Gemini 3.5 Flash-LiteGoogle0.0%0.0%0.0%0.0%
Gemini 3.1 Flash-Lite (minimal reasoning)Google0.0%8.3%0.2%0.0%
Gemini Decision 1Google0.0%0.0%0.0%0.0%
Gemini 3.5 Flash (minimal reasoning)Google33.3%25.0%0.6%0.5%
Clef OmniCloudflare0.0%0.0%0.0%0.0%
ReflexMorph0.0%0.0%0.0%0.0%
Elephant 1Baseten0.0%0.0%0.0%0.0%
Mercury DecideInception0.0%0.0%0.0%0.0%
Drex Max FinanceNace AI0.0%0.0%0.0%0.0%
Span-02Respan0.0%0.0%0.0%0.0%
GPT-6.1 Sol Ultrafast (low reasoning)OpenAI50.0%25.0%0.0%0.0%
d1.1Liquid AI0.0%0.0%0.0%0.0%
d1Liquid AI0.0%0.0%0.0%0.0%
Decider v1.2Perplexity0.0%0.0%0.0%0.0%
Mammoth 1Baseten0.0%0.0%0.0%0.0%
Seed-2.0-Mini (minimal reasoning)ByteDance0.0%0.0%0.8%0.4%
Mistral Small 4 (no reasoning)Mistral AI0.0%25.0%1.0%0.6%
Drex v2.0Nace AI0.0%0.0%0.0%0.0%
GPT-5.4 mini (no reasoning)OpenAI0.0%50.0%0.0%0.0%
Inkling Small (no reasoning)Thinking Machines0.0%0.0%5.3%7.2%
Jev 1.13TypeSafe0.0%0.0%0.0%0.0%
Mercury 3 (no reasoning)Inception0.0%0.0%0.4%0.0%
Joey 1Baseten0.0%0.0%0.0%0.0%
Clef FlashCloudflare0.0%0.0%0.0%0.0%
ClefCloudflare0.0%0.0%0.0%0.0%
d1 UltraLiquid AI0.0%0.0%0.0%0.0%
Mistral Medium 3.5 (no reasoning)Mistral AI0.0%0.0%0.0%0.4%
Mistral Medium 3.5Mistral AI0.0%8.3%0.3%0.8%
GPT-6 Luna DecisionsOpenAI0.0%0.0%0.0%0.0%
Parallel Decisions FastParallel0.0%0.0%0.0%0.0%
Jev 1.14TypeSafe0.0%0.0%0.0%0.0%
GPT-6.1 Terra (low reasoning)OpenAI62.5%75.0%0.0%0.0%
Drex v1.5Nace AI0.0%8.3%0.0%0.0%
Decider Flash v1.2Perplexity0.0%0.0%0.0%0.0%
Microsoft Decision 1Microsoft0.0%0.0%0.0%0.0%
Decider V1 27BPerplexity0.0%8.3%0.0%0.0%
Drex MaxNace AI25.0%33.3%0.0%0.0%
GLM-5.3 FlashX Legal (low reasoning)Z.ai62.5%37.5%0.0%0.0%
Mistral Large 4Mistral AI58.3%–1.8%–
Gemini 3.8 Flash (low reasoning)Google83.3%50.0%0.0%0.0%
GPT-5.4 nano (no reasoning)OpenAI75.0%66.7%4.0%3.1%
GPT-6.1 Sol UltrafastOpenAI83.3%41.7%0.0%0.0%
Claude Haiku 4.5Anthropic83.3%25.0%0.0%0.3%
Claude Sonnet 5.5 (low reasoning)Anthropic91.7%–0.0%–
Gemini 3.6 Flash (minimal reasoning)Google91.7%41.7%0.0%0.3%
Kev 4BJared Palmer66.7%–0.0%–

Lightning: strength against cost

What one Lightning game cost at the run date's prices, on a log scale. Up and to the left is better; the dashed line joins the strongest model at each cost.

Named: the six best and the best for the moneyOther models (hover for names)Best score at each cost

-200300800 $0.0001$0.001$0.01$0.1$1 Cost per game, US dollars (log scale) Ladder Elo Gemini 3.5 Flash-Lite Gemini 3.1 Flash-Lite (minimal reasoning) Gemini Decision 1 Gemini 3.5 Flash (minimal reasoning) Clef Omni Reflex Span-02

Best per budget

  1. Span-02: 614 at $0.0001
  2. Reflex: 616 at $0.0007
  3. Gemini Decision 1: 755 at $0.0018
  4. Gemini 3.1 Flash-Lite (minimal reasoning): 780 at $0.0058
  5. Gemini 3.5 Flash-Lite: 892 at $0.0070

Cheapest first: each model here beats every cheaper one on score.

How BulletBench works

  1. 1

    The position

    The model gets the position, the moves so far, its remaining clock and the legal moves.

  2. 2

    The reply

    It answers with one move. The whole API round trip, including any hidden reasoning, comes off its clock.

  3. 3

    The engine

    Stockfish replies instantly at the model's current ladder level.

  4. 4

    The ladder

    Win and the next game is a level up; lose and it is a level down; draw and it stays. The results give a rating with a 95% interval.

Why chess on a clock

Many products need an answer in about a second: routing, triage, autocomplete, live agents. BulletBench asks which models can still think usefully at that speed. Chess gives a hard, objective score, and the clock makes slow thinking a real cost instead of a free extra.

The clocks

Lightning 10+1
Ten seconds, plus one second a move. A model can play for ever, but only at about a second an answer.
Bullet 60s
One minute for the whole game, no increment.
Blitz 3+2
Three minutes plus two seconds a move: room for reasoning models, for comparison.

Fair play

Same prompt
Every model gets the same system prompt and position format.
Invalid replies
A reply with no legal move gets two corrective retries, then a random legal move is played and counted.
Settings
Each model runs at its provider's default and, where the provider offers one, its fastest reasoning setting, shown separately.

Honest caveats

A dozen games or fewer per clock (the table gives each configuration's count) give wide intervals, so neighbouring ranks are often within each other's ranges. Ratings also come in steps: the engine plays at fixed levels and the rating is fitted from which levels a model beat, drew or lost to, so models with the same results against the same levels get exactly the same rating. Latency depends on the provider's servers on the day. The ladder's Elo anchors come from v1's Stockfish; this edition runs Stockfish 19, so ratings are not comparable with v1's.

What it measures

  • Decision quality under real time pressure
  • Response latency at the provider's default settings
  • Which reasoning settings are fast enough to use

What it does not measure

  • Any business skill: this is chess against an engine
  • Chess strength without a clock

Method

  • Games against a Stockfish ladder; every second of response time comes off the model's clock
  • Each model at its default setting and, where offered, its fastest reasoning setting
  • Ladder Elo with a 95% interval; it orders models, it is not a FIDE rating

How all our benchmarks work

Use the data

Every result on this page is free to reuse under CC BY 4.0: results.csv · results.json. Credit Spring Prompt with a link. All downloads · Embed the chart

Embed this chart

Put BulletBench's current top 10 in your article. Free to reuse under CC BY 4.0; the code includes the credit link. Add ?n=5 for fewer rows, ?models= with model slugs (as in our model page addresses) for your own comparison, or ?theme=dark.

Interactive chart

<iframe src="https://springprompt.com/embed/bulletbench" title="BulletBench results by Spring Prompt" width="100%" height="456" style="border:0;max-width:720px" loading="lazy"></iframe>
<p><a href="https://springprompt.com/benchmarks/bulletbench">BulletBench results by Spring Prompt</a></p>

Image, for sites that block iframes

<a href="https://springprompt.com/benchmarks/bulletbench"><img src="https://springprompt.com/og/benchmarks/bulletbench.png" alt="BulletBench: the top 10 AI models by ladder elo" width="1200" height="630" style="max-width:100%;height:auto"></a>
<p><a href="https://springprompt.com/benchmarks/bulletbench">BulletBench results by Spring Prompt</a></p>
See the image