Best LLMs for Instruction Following — October 2026 Leaderboard
Data refreshed:
Ability to follow precise instructions and constraints
As of October 2026, the top instruction following model on the BenchLM leaderboard is Grok 4.3 with a weighted instruction following score of 97.1.
Decision lens: use provisional-ranked mode for broader public evidence and verified-ranked mode for source-only comparisons. A model can move between views as evidence coverage changes.
- Data refreshed
- October 20, 2026
- Provisional-ranked
- 148 of 959 models
- Verified-ranked
- 156 of 959 models
- Weighted evidence
- 2 of 5 benchmarks
5 tracked benchmarks
IFEval, IFBench, SOB Value Acc, Gray Swan IPI (15 attempts)
Evidence set: IFEval, IFBench, SOB Value Acc, Gray Swan IPI (15 attempts)
Best Instruction Following picks
BenchLM summaries for instruction following plus the practical tradeoffs users check next: open weights, price, speed, latency, and context.
IFEval Leaderboard
Primary score: weighted instruction following score. Higher values rank first. Use the Show metric control to change the value shown in each row.
Switch between provisional-ranked and verified-ranked modes to compare the broader public dataset with sourced-only rankings.
Filters
| Parameters (B) | ||||||||
|---|---|---|---|---|---|---|---|---|
1 Grok 4.3 xAI | Not reported | Closed | 97.1% | 68 | — | 81.3% | — | — |
2 | Not reported | Open | 94.7% | 81 | — | 82.8% | — | — |
3 MAI-Thinking-1 Microsoft | Not reported | Closed | 94.7% | 55 | — | 85% | — | — |
4 GPT-5.2-Codex OpenAI | Not reported | Closed | 93.8% | 75 | 92%P | — | — | — |
5 | Not reported | Open | 93.8% | 65 | — | — | — | — |
| Not reported | Open | 93.8% | 64 | 92.6%P | — | — | — | |
7 MiMo-V2.5-Pro Xiaomi | Not reported | Closed | 93.8% | 68 | — | — | — | — |
8 GPT-5.5 OpenAI | Not reported | Closed | 93.3% | 74 | — | — | — | — |
9 | Not reported | Open | 93.3% | 64 | — | 82.2% | — | — |
10 Muse Spark Meta | Not reported | Closed | 93.3% | 63 | — | — | — | — |
11 GPT-5.4 nano OpenAI | Not reported | Closed | 93.3% | 50 | — | — | — | — |
12 | Not reported | Open | 93% | 61 | 93.4% | — | — | — |
13 | Not reported | Open | 93% | 60 | — | 75.7%P | — | — |
14 | Not reported | Open | 92.9% | 59 | 95% | — | — | — |
15 | Not reported | Open | 92.9% | 56 | — | — | — | — |
16 Qwen3.7 Max Alibaba | Not reported | Closed | 92.8% | 76 | 94.3% | 79.1% | — | — |
17 Qwen3.7 Plus Alibaba | Not reported | Closed | 92.8% | 71 | 94.6% | 79.1% | — | — |
18 GPT-5.3 Codex OpenAI | Not reported | Closed | 92.6% | 75 | 93%P | 78.1%P | — | — |
19 GPT-5.2 OpenAI | Not reported | Closed | 92.6% | 58 | 94%P | 76.2%P | — | — |
20 | Not reported | Open | 92.1% | 57 | — | 81.7% | — | — |
21 Qwen3.8-Omni-Flash Alibaba | Not reported | Closed | 91.6% | 69 | — | 81.5% | — | — |
22 | Not reported | Open | 91.2% | 64 | — | 81.3% | — | — |
23 GPT-5.4 OpenAI | Not reported | Closed | 90.7% | 70 | 96%P | 79.4%P | — | — |
24 | Not reported | Open | 90.7% | 53 | — | — | — | — |
25 | Not reported | Open | 90.1% | 59 | — | — | — | — |
Top AI models for Instruction Following — October 2026
As of October 2026, Grok 4.3 leads the provisional instruction following leaderboard with a score of 97.1%, followed by Qwen3.8 Max (94.7%) and MAI-Thinking-1 (94.7%). BenchLM is currently showing 148 provisional-ranked models and 156 verified-ranked models in this category.
Scored on 2 of the 2 weighted instruction following benchmarks.
Scored on 1 of the 2 weighted instruction following benchmarks.
Scored on 1 of the 2 weighted instruction following benchmarks.
What changed
Grok 4.3 ranks #1 at 97.1 on the weighted instruction following score.
Qwen3.8 Max ranks #2 at 94.7 on the weighted instruction following score.
MAI-Thinking-1 ranks #3 at 94.7 on the weighted instruction following score.
Top models by benchmark
Instruction-following benchmark used in first-party comparison charts for agent-oriented reasoning models.(70% of category score)
Score in Context
What these scores mean
The Instruction following leaderboard ranks models by a weighted category score. The overall ranking combines external indices with benchmark evidence rather than fixed category weights. It directly measures reliability: a model that ignores formatting rules, word count limits, or inclusion constraints is unusable in automated pipelines. The weighted score is led by IFBench at 70% and AA-IFBench at 30%.
Known limitations
IFEval tests a specific set of verifiable constraints (word count, formatting, inclusion/exclusion) — it doesn't capture the full range of instruction-following quality. A model can score well on IFEval but still misinterpret nuanced or ambiguous instructions. IFBench is newer and coverage is still building.
How we weight
The Instruction following leaderboard ranks models by a weighted category score. The overall ranking combines external indices with benchmark evidence rather than fixed category weights. It directly measures reliability — a model that ignores constraints is unusable in production pipelines. See the instruction following leaderboard.
Leaderboards exclude benchmark rows that BenchLM generated from other scores or cloned from reference models. When a weighted benchmark is missing after that filter, the category falls back to the remaining trustworthy public rows instead of filling the gap with synthetic values.
The full scoring rules, freshness handling, and runtime/pricing caveats live on the BenchLM methodology page.
Scroll horizontally to read the full evidence ledger.
| Benchmark | Weight | Status | Description |
|---|---|---|---|
| IFBench | 70% | Weighted | Instruction-following benchmark used in first-party comparison charts for agent-oriented reasoning models. |
| AA-IFBench | 30% | Weighted | |
| IFEval | — | Display only | Tests ability to follow verifiable instructions like format constraints and content requirements |
| SOB Value Acc | — | Display only | A structured-output benchmark from Interfaze measuring whether extracted JSON leaf values exactly match verified ground truth. |
| Gray Swan IPI (15 attempts) | — | Display only | Measures how often indirect prompt-injection attacks succeed within fifteen attempts. Lower scores mean greater resistance to the tested attacks. |
About Instruction Following benchmarks
Tests ability to follow verifiable instructions like format constraints and content requirements
Questions
What is the best LLM for instruction following?
The best instruction-following LLMs are ranked by benchmarks like IFEval, which measure how accurately models adhere to specific formatting, length, and content constraints.
What is IFEval and how does it work?
IFEval (Instruction Following Evaluation) tests whether LLMs can precisely follow verifiable instructions such as word count limits, formatting rules, and inclusion/exclusion constraints.
Why is instruction following important for LLMs?
Instruction following is critical because real-world applications require models to reliably adhere to user specifications, output formats, and constraints rather than just generating plausible text.
Instruction following updates
Instruction following scores just got interesting. Get the weekly update.
Read a sample issueJoin 5,600+ readers.
One email each week. Unsubscribe anytime.