Benchmark Almanac

IFScale

Gemini 2.5 Pro Jun 2025 68.9% leads since July 2025
Not in model cards

Models write business reports while including up to 500 specified keywords. The benchmark measures how many instructions they follow at each density.

The authors report 20 models from seven providers. Arize independently repeated the evaluation on newer models.

Made by
Distyl AI
Released
July 2025
Kind
Following precise instructions
Level
University
Scored
Checked automatically
Answers
Open-ended
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Documents
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
In views
Law and finance
Strongest model tested
o3 #84 on the capability index
Last model added
September 2025
Scores refreshed
November 21st, 2026
22
Entries in its paper, 20 models we track

Gemini 2.5 Pro leads with 68.9% among models released by September 2025 Gemini 2.5 Pro leads with 68.9% among models released by September 2025

Best score per entry from the benchmark's paper, read on November 21st, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in September 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Gemini 2.5 Pro Jun 2025 68.9%
  2. 2 o3 high 62.8%
  3. 3 Grok 3 61.9%
  4. 4 Claude 3.7 Sonnet 52.7%
  5. 5 claude-opus-4 (r) 52.1%
  6. 6 GPT-4.1 48.9%
  7. 7 Claude Opus 4 44.6%
  8. 8 GPT-4.5 43%
  9. 9 Claude Sonnet 4 42.9%
  10. 10 claude-sonnet-4 (r) 39.9%
  11. 11 Grok-3 mini 36.4%
  12. 12 o4-mini medium 34.4%
  13. 13 Gemini 2.5 Flash Sep 2025 34.2%
  14. 14 DeepSeek-R1 May 2025 30.9%
  15. 15 Llama 4 Maverick 27.7%
  16. 16 GPT-4.1 mini 27.2%
  17. 17 Qwen3-235B-A22B 20.9%
  18. 18 GPT-4.1 nano 16.2%
  19. 19 GPT-4o Mar 2025 15.4%
  20. 20 GPT-4o mini 10.4%
  21. 21 Claude 3.5 Haiku 8.50%
  22. 22 Llama 4 Scout 6.70%

Source: The IFScale paper

#EntryScoreSettingDateSource
1 Gemini 2.5 Pro Jun 2025 68.9% Jun 2025 Official leaderboard
2 o3 62.8% high Apr 2025 Official leaderboard
3 Grok 3 61.9% Apr 2025 Official leaderboard
4 Claude 3.7 Sonnet 52.7% Feb 2025 Official leaderboard
5 claude-opus-4 (r) 52.1% Official leaderboard
6 GPT-4.1 48.9% Apr 2025 Official leaderboard
7 Claude Opus 4 44.6% May 2025 Official leaderboard
8 GPT-4.5 43% Feb 2025 Official leaderboard
9 Claude Sonnet 4 42.9% May 2025 Official leaderboard
10 claude-sonnet-4 (r) 39.9% Official leaderboard
11 Grok-3 mini 36.4% Apr 2025 Official leaderboard
12 o4-mini 34.4% medium Apr 2025 Official leaderboard
13 Gemini 2.5 Flash Sep 2025 34.2% Sep 2025 Official leaderboard
14 DeepSeek-R1 May 2025 30.9% May 2025 Official leaderboard
15 Llama 4 Maverick 27.7% Apr 2025 Official leaderboard
16 GPT-4.1 mini 27.2% Apr 2025 Official leaderboard
17 Qwen3-235B-A22B 20.9% Apr 2025 Official leaderboard
18 GPT-4.1 nano 16.2% Apr 2025 Official leaderboard
19 GPT-4o Mar 2025 15.4% Mar 2025 Official leaderboard
20 GPT-4o mini 10.4% Jul 2024 Official leaderboard
21 Claude 3.5 Haiku 8.50% Oct 2024 Official leaderboard
22 Llama 4 Scout 6.70% Apr 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Gemini 2.5 Pro Jun 2025 68.9% Jun 2025 Official leaderboard
2 o3 62.8% high Apr 2025 Official leaderboard
3 Grok 3 61.9% Apr 2025 Official leaderboard
4 Claude 3.7 Sonnet 52.7% Feb 2025 Official leaderboard
5 GPT-4.1 48.9% Apr 2025 Official leaderboard
6 Claude Opus 4 44.6% May 2025 Official leaderboard
7 GPT-4.5 43% Feb 2025 Official leaderboard
8 Claude Sonnet 4 42.9% May 2025 Official leaderboard
9 Grok-3 mini 36.4% Apr 2025 Official leaderboard
10 o4-mini 34.4% medium Apr 2025 Official leaderboard
11 Gemini 2.5 Flash Sep 2025 34.2% Sep 2025 Official leaderboard
12 DeepSeek-R1 May 2025 30.9% May 2025 Official leaderboard
13 Llama 4 Maverick 27.7% Apr 2025 Official leaderboard
14 GPT-4.1 mini 27.2% Apr 2025 Official leaderboard
15 Qwen3-235B-A22B 20.9% Apr 2025 Official leaderboard
16 GPT-4.1 nano 16.2% Apr 2025 Official leaderboard
17 GPT-4o Mar 2025 15.4% Mar 2025 Official leaderboard
18 GPT-4o mini 10.4% Jul 2024 Official leaderboard
19 Claude 3.5 Haiku 8.50% Oct 2024 Official leaderboard
20 Llama 4 Scout 6.70% Apr 2025 Official leaderboard

Rankings by part

We read 4 parts from the benchmark’s paper, each ranking models on one part of the test, such as 10 instructions, 50 instructions and 100 instructions. Grok 3 leads 2 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

10 instructions

20 models we track have a score here.

  1. 1 Claude 3.7 Sonnet ↑3 100%
  2. 2 GPT-4.5 ↑5 100%
  3. 3 Llama 4 Maverick ↑10 100%
  4. 4 Llama 4 Scout ↑16 100%
  5. 5 Grok 3 ↓2 100%
  6. 6 Grok-3 mini ↑3 100%
  7. 7 GPT-4.1 mini ↑7 100%
  8. 8 o3 ↓6 100%
  9. 9 o4-mini ↑1 100%
  10. 10 Qwen3-235B-A22B ↑5 100%

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Google DeepMind
Benchmark paper Not tested since Sep 2025

Source: The IFScale paper

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) Gemini 2.5 Pro o3 Claude 3.7 Sonnet GPT-4o mini 0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) Gemini 2.5 Pro Grok 3

Source: Official IFScale leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Instructions and chat

All 144 in Instructions and chat
  • IFEval In use Models answer about 500 prompts containing verifiable instructions such as word limits or required keywords. Code checks each instruction for strict and loose accuracy. Gemma 4 31B IT 98.9%
  • IFBench In use 58 new out-of-domain verifiable output constraints. Scored by programmatic constraint checks (strict or loose). Flint 89.2%
  • MultiChallenge In use Multi-turn conversations testing instruction retention, inference memory, versioned editing and self-coherence. Graded by LLM judge with rubrics. GPT-6.1 Sol 82%
  • Multi-IF In use Models follow verifiable instructions over three-turn conversations in eight languages, extending IFEval. Scored by instruction accuracy at each turn. LFM3-1.7B 89.1%
  • AdvancedIF In use Models respond to over 1,600 prompts with complex, multi-turn, and system-level instructions. A judge model checks each response against expert-written rubrics. MAI-Thinking-1 85%
  • MM-IFEval In use Multimodal models follow image-grounded instructions with exact output constraints. Checks mix rule-based verification and a judge model. GPT-4o 64.6%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (22 entries).
  • Pass Models scored 20 models we track, plus 2 other entries on its leaderboard.
  • Warning Newest score The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash.
  • Note Frontier models None of the top 10 has a score. The strongest tested is o3 (#84).
  • Pass Headroom The best is 68.9%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on distylai.github.io: 21 rows for models we track and 2 other entries at the last read.
  • Pass Last read November 21st, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check