Capability ranking
Best AI for Instruction Following
Best Instruction-Following Models
Updated 2026-08-11
Ranked models50
Vendors10
Open-weight models13
Official prices23
What this ranking says
A quick reading of the current instruction ranking before the full data table.
Nova 2 Pro leads this page at 94.7, 2.5 points ahead of Qwen3.7 Max.
Nemotron 3 Ultra is the highest-ranked open-weight option at #4. GPT-4.1-nano has the lowest official input price among ranked models.
Instruction Ranking Leaders
The leading instruction scores, with the axis focused on the competitive range rather than zero.
Instruction Capability Profile
Compare instruction strength with each model's broader capability score.
ModelReasoningKnowledgeMathCodingInstruction
Instruction scoreOverall score
50 models with comparable dataOpen a model by selecting its logo
Instruction Model Ranking
Models are ordered by their instruction score.
| 01 | AM Nova 2 ProAmazon | Nova 2 Pro | 94.7 | 52.5 | 66.7 | 68.0 | 68.6 | 67.8 | 100.0% | N/A | Closed weights | N/A | N/A |
| 02 | AC | Qwen3.7 Max | 92.2 | 78.8 | 91.5 | 89.1 | 93.1 | 85.6 | 100.0% | 1M | Closed weights | $2.5 | $7.5 |
| 03 | AC | Qwen3.7-Plus | 87.8 | 73.8 | 84.7 | 83.5 | 66.9 | 61.6 | 100.0% | 1M | Closed weights | $0.50 | $3 |
| 04 | NV | Nemotron 3 Ultra (550B A55B) | 85.3 | 62.9 | 82.6 | 77.3 | 100.0 | 49.3 | 100.0% | 1M | Open weights | $0.50 | $2.5 |
| 05 | AC | Qwen3-235B-A22B-Thinking-2507 | 84.9 | 48.5 | 71.4 | 61.9 | 63.5 | 53.9 | 100.0% | 262.1K | Closed weights | N/A | N/A |
| 06 | NR | Hermes 3 70B | 82.8 | 7.2 | 59.2 | 39.6 | 1.4 | N/A | 80.0% | 131.1K | Closed weights | N/A | N/A |
| 07 | AC | Qwen3.5-397B-A17B | 81.9 | 66.3 | 81.3 | 82.8 | 50.1 | 57.4 | 100.0% | 262.1K | Open weights | $0.60 | $3.6 |
| 08 | AC | Qwen3 VL 235B A22B Thinking | 77.7 | 45.8 | 58.6 | 55.9 | 53.8 | 46.1 | 100.0% | 262.1K | Closed weights | N/A | N/A |
| 09 | AC | Qwen3 VL 235B A22B Instruct | 76.8 | 43.2 | 54.0 | 64.8 | 28.0 | 53.2 | 100.0% | 262.1K | Closed weights | N/A | N/A |
| 10 | AC | Qwen3-235B-A22B-Instruct-2507 | 72.8 | 42.4 | 61.1 | 70.6 | 21.5 | 51.2 | 100.0% | 262.1K | Open weights | $0.70 | $2.8 |
| 11 | AC | Qwen3.6 Plus | 72.7 | 67.5 | 80.9 | 87.3 | 73.3 | 51.8 | 100.0% | 1M | Closed weights | $0.50 | $3 |
| 12 | NV | Nemotron 3 Super (120B A12B) | 71.4 | 44.8 | 70.0 | 60.8 | 68.7 | 31.9 | 100.0% | 262.1K | Open weights | $0.20 | $0.80 |
| 13 | AMNova 2 LiteAmazon | Nova 2 Lite | 71.4 | 43.0 | 62.0 | 63.3 | 62.0 | 45.6 | 100.0% | 1M | Closed weights | $0.33 | $2.75 |
| 14 | AC | Qwen3.5-27B | 70.6 | 57.2 | 62.5 | 78.0 | 59.4 | 47.9 | 100.0% | 262.1K | Open weights | $0.30 | $2.4 |
| 15 | OP | o3-mini | 70.0 | 36.3 | 58.1 | 46.9 | 63.8 | 51.7 | 100.0% | 200K | Closed weights | $1.1 | $4.4 |
| 16 | AC | Qwen3.5-122B-A10B | 69.8 | 59.2 | 67.4 | 84.6 | 56.3 | 54.2 | 100.0% | 262.1K | Open weights | $0.40 | $3.2 |
| 17 | AC | Qwen2.5 72B Instruct | 68.4 | 21.4 | 22.5 | 37.5 | 80.1 | 60.3 | 100.0% | 131.1K | Closed weights | $1.4 | $5.6 |
| 18 | CO | Command A+ | 64.9 | 41.9 | N/A | N/A | 58.4 | 19.4 | 60.0% | 128K | Open weights | $2.5 | $10 |
| 19 | MA | Kimi K2-Instruct-0905 | 63.4 | 37.2 | 55.1 | 38.0 | 34.5 | 35.3 | 100.0% | N/A | Closed weights | $0.60 | $2.5 |
| 20 | AC | Qwen3 VL 32B Thinking | 63.4 | 40.5 | 47.9 | 66.5 | 45.1 | 33.6 | 100.0% | N/A | Closed weights | N/A | N/A |
| 21 | AC | Qwen3-Next-80B-A3B-Thinking | 62.7 | 40.9 | 58.2 | 65.2 | 45.7 | 38.5 | 100.0% | 131.1K | Open weights | $0.50 | $6 |
| 22 | AC | Qwen3-Next-80B-A3B-Instruct | 62.7 | 36.6 | 53.4 | 53.0 | 18.9 | 49.0 | 100.0% | 131.1K | Open weights | $0.50 | $2 |
| 23 | GO | Gemma 3 27B | 61.1 | 15.8 | 46.8 | 37.9 | 89.0 | 39.2 | 100.0% | 131.1K | Closed weights | N/A | N/A |
| 24 | AMNova 2 OmniAmazon | Nova 2 Omni | 58.9 | 43.9 | N/A | 62.1 | 67.3 | N/A | 60.0% | N/A | Closed weights | N/A | N/A |
| 25 | NV | Nemotron 3 Nano (30B A3B) | 56.7 | 35.2 | 54.3 | 40.2 | 92.0 | 12.1 | 100.0% | 262.1K | Open weights | N/A | N/A |
| 26 | AC | Qwen3.5-35B-A3B | 54.4 | 52.0 | 61.8 | 71.2 | 43.8 | 44.2 | 100.0% | 262.1K | Open weights | $0.25 | $2 |
| 27 | GO | Gemma 3 12B | 51.4 | 11.2 | 39.4 | 35.4 | 74.8 | 32.4 | 100.0% | 131.1K | Closed weights | N/A | N/A |
| 28 | GO | Gemma 3 4B | 50.4 | 0.0 | 21.7 | 13.3 | 51.8 | 10.8 | 100.0% | 131.1K | Closed weights | N/A | N/A |
| 29 | AC | Qwen3.5-9B | 44.6 | 41.4 | 47.6 | 52.0 | 28.1 | 33.6 | 100.0% | 262.1K | Open weights | N/A | N/A |
| 30 | AC | Qwen3 VL 8B Thinking | 41.3 | 28.5 | 32.5 | 47.0 | 33.6 | 28.9 | 100.0% | 262.1K | Closed weights | N/A | N/A |
| 31 | AC | Qwen3 VL 32B Instruct | 41.0 | 35.2 | 39.3 | 43.3 | 19.5 | 9.6 | 100.0% | N/A | Closed weights | N/A | N/A |
| 32 | AC | Qwen3 VL 30B A3B Thinking | 39.9 | 31.8 | 37.8 | 43.4 | 38.8 | 30.8 | 100.0% | 262.1K | Closed weights | N/A | N/A |
| 33 | OP | GPT-4.5 | 38.7 | 36.2 | 43.8 | 91.0 | 48.7 | 33.4 | 100.0% | 128K | Closed weights | N/A | N/A |
| 34 | MI | MAI-Thinking-1 | 38.4 | 55.1 | 74.5 | 71.1 | 58.8 | 45.8 | 100.0% | N/A | Closed weights | N/A | N/A |
| 35 | AC | Qwen3 VL 30B A3B Instruct | 37.6 | 28.8 | 30.6 | 41.6 | 14.3 | 7.7 | 100.0% | 262.1K | Closed weights | N/A | N/A |
| 36 | AC | Qwen3.5-4B | 36.3 | 32.4 | 35.8 | 36.9 | 18.8 | 25.0 | 100.0% | N/A | Closed weights | N/A | N/A |
| 37 | AC | Qwen3 VL 8B Instruct | 36.0 | 24.8 | 22.4 | 25.4 | 5.2 | 3.9 | 100.0% | 262.1K | Closed weights | N/A | N/A |
| 38 | AC | Qwen2.5 7B Instruct | 35.0 | 6.5 | 9.4 | 17.2 | 58.4 | 37.6 | 100.0% | 131.1K | Closed weights | $0.175 | $0.70 |
| 39 | MA | Mistral Small 3 24B Instruct | 32.2 | 7.5 | 18.0 | 23.4 | 48.6 | 43.9 | 100.0% | 32K | Closed weights | N/A | N/A |
| 40 | OP | GPT-4.1 | 29.3 | 31.9 | 38.6 | 49.0 | 3.5 | 32.3 | 100.0% | 1M | Closed weights | $2 | $8 |
| 41 | AC | Qwen3 VL 4B Thinking | 29.2 | 23.0 | 28.4 | 30.2 | 20.9 | 13.5 | 100.0% | 262.1K | Closed weights | N/A | N/A |
| 42 | OP | GPT-4.1 mini | 21.1 | 24.8 | 35.6 | 38.6 | 5.3 | 13.1 | 100.0% | 1M | Closed weights | $0.40 | $1.6 |
| 43 | NV | Llama 3.1 Nemotron Nano 8B V1 | 17.7 | 10.4 | 28.8 | N/A | 27.0 | 84.4 | 80.0% | N/A | Closed weights | N/A | N/A |
| 44 | OP | GPT-4o | 17.4 | 23.9 | 22.8 | 35.2 | 0.0 | 8.5 | 100.0% | 128K | Closed weights | $2.5 | $10 |
| 45 | AC | Qwen3 VL 4B Instruct | 16.8 | 19.3 | 12.3 | 33.4 | 4.0 | 1.9 | 100.0% | 262.1K | Closed weights | N/A | N/A |
| 46 | GO | Gemma 3 1B | 11.7 | 0.0 | 0.1 | 0.6 | 6.9 | 1.0 | 100.0% | N/A | Closed weights | N/A | N/A |
| 47 | AC | Qwen3.5-2B | 11.3 | 2.8 | 15.1 | 13.3 | N/A | N/A | 60.0% | N/A | Closed weights | N/A | N/A |
| 48 | MA | Pixtral-12B | 5.3 | 0.0 | N/A | 14.1 | 18.6 | 16.9 | 80.0% | 128K | Open weights | $0.15 | $0.15 |
| 49 | OP | GPT-4.1 nano | 2.6 | 4.3 | 24.0 | 38.4 | 1.9 | 0.0 | 100.0% | 1M | Closed weights | $0.10 | $0.40 |
| 50 | AC | Qwen3.5-0.8B | 0.6 | 0.0 | 2.2 | 3.3 | N/A | N/A | 60.0% | N/A | Closed weights | N/A | N/A |
Official PAYG prices appear here. Third-party offers remain on the pricing and model detail pages.
Capability at each price point
Official vendor API prices plotted against the instruction metric used on this page. Price remains a separate decision signal.
Instruction scoreOfficial API price blend (8:1 input/output), USD / 1M
Efficient frontier23 models with official PAYG prices
Instruction Ranking Composition
Vendor concentration and model access among the first 25 ranked products.
Top-model vendor mix
25 modelsCompare the leaders
Open focused comparisons between the current leader and the nearest practical alternatives.
About this leaderboard
Common questions about Best AI for Instruction Following.
What is the best model on Best AI for Instruction Following?
Nova 2 Pro is currently ranked first with a instruction score of 94.7.
How is this leaderboard ranked?
Each model appears once using its current scored version. Category pages use the capability named in the title, while the overall page uses the LLMBoard score.
How many models are included?
This page currently ranks 50 unique model products.
Where does pricing come from?
Main price columns use the model vendor's official standard PAYG API rate. Eligible third-party offers appear only in separately labeled columns, and unavailable official prices display as N/A.
Does Arena affect the score?
No. Arena results are displayed as an independent signal and are not included in the current LLMBoard capability score.