/best/researchupdated May 202685 models evaluated

Best AI for research

Synthesizing sources, answering with citations, deep reading.

CodingMathWritingResearchTranslationData AnalysisChatbotsStudentsBusinessCreative WritingTabular Data & Spreadsheets

Research workloads punish hallucination harder than almost any other task. The model has to read long, contradictory sources and produce an answer that a specialist wouldn't laugh at. Long-context retention and tool-use (search, browsing) are table stakes — a model with a 128K context window that loses track of page 1 by page 100 is useless for real research.

What matters: long-context faithfulness, reasoning chain quality, and willingness to say 'I don't know' rather than confabulate. Models with retrieval-augmented architectures (like Sonar Pro) have a natural edge here, but frontier chat models with large context windows are competitive when source material is provided directly.

Our research rank weights long-context (2×), reasoning (1.5×), tool-use (1×), and multilingual (0.5×). Multilingual matters more than people expect — research sources are rarely all in English.

Full rankings

All 85 models, scored for research

weighted composite · lower-is-worse
#ModelProviderTask score$/in$/outContext
01Qwen: Qwen3 235B A22B Instruct 2507QQwen5.00/5.0$0.071$0.100262K
02OpenAI: gpt-oss-120bOOpenAI5.00/5.0$0.039$0.180131K
03NVIDIA: Nemotron 3 Nano 30B A3BNNVIDIA5.00/5.0$0.050$0.200262K
04Xiaomi: MiMo-V2-FlashXXiaomi5.00/5.0$0.100$0.300262K
05Gemma 4 26B A4B GGoogle5.00/5.0$0.060$0.330262K
06GLM-4.7 FlashZZhipu AI5.00/5.0$0.060$0.400203K
07DeepSeek V3.2DDeepSeek5.00/5.0$0.252$0.378131K
08NVIDIA: Nemotron 3 SuperNNVIDIA5.00/5.0$0.090$0.4501M
09Grok 4.1 FastXXAI5.00/5.0$0.200$0.5002M
10Qwen: Qwen3.5-35B-A3BQQwen5.00/5.0$0.139$1.00262K
11Qwen: Qwen3.6 35B A3BQQwen5.00/5.0$0.150$1.00262K
12GLM-4.7ZZhipu AI5.00/5.0$0.400$1.75203K
13Qwen: Qwen3.5 Plus 2026-04-20QQwen5.00/5.0$0.300$1.801M
14Qwen: Qwen3.6 PlusQQwen5.00/5.0$0.325$1.951M
15GPT-5 MiniOOpenAI5.00/5.0$0.250$2.00400K
16Qwen: Qwen3.6 27BQQwen5.00/5.0$0.300$2.00262K
17MiMo-V2.5XXiaomi5.00/5.0$0.400$2.001.0M
18xAI: Grok Build 0.1XXAI5.00/5.0$1.00$2.00256K
19Grok 4.20XXAI5.00/5.0$1.25$2.502M
20Gemini 3 Flash PreviewGGoogle5.00/5.0$0.500$3.001.0M
21MiMo-V2.5-ProXXiaomi5.00/5.0$1.00$3.001.0M
22MoonshotAI: Kimi K2.6MMoonshotAI5.00/5.0$0.730$3.49262K
23GPT-5.4 MiniOOpenAI5.00/5.0$0.750$4.50400K
24o4 MiniOOpenAI5.00/5.0$1.10$4.40200K
25Claude Haiku 4.5AAnthropic5.00/5.0$1.00$5.00200K
26Qwen: Qwen3.6 Max PreviewQQwen5.00/5.0$1.04$6.24262K
27Qwen 3.7 MaxQQwen5.00/5.0$2.50$7.501M
28GPT-4.1OOpenAI5.00/5.0$2.00$8.001.0M
29GPT-5OOpenAI5.00/5.0$1.25$10.00400K
30GPT-5.1OOpenAI5.00/5.0$1.25$10.00400K
31Gemini 3.1 Pro PreviewGGoogle5.00/5.0$2.00$12.001.0M
32GPT-5.2OOpenAI5.00/5.0$1.75$14.00400K
33GPT-5.4OOpenAI5.00/5.0$2.50$15.001.1M
34Claude Sonnet 4.6AAnthropic5.00/5.0$3.00$15.001M
35Grok 4XXAI5.00/5.0$3.00$15.00256K
36Grok 3XXAI5.00/5.0$3.00$15.00131K
37Claude Opus 4.7AAnthropic5.00/5.0$5.00$25.001M
38Claude Opus 4.6AAnthropic5.00/5.0$5.00$25.001M
39DeepSeek V4 FlashDDeepSeek4.67/5.0$0.100$0.2001.0M
40Gemma 4 31BGGoogle4.67/5.0$0.120$0.370262K
41DeepSeek V3.1DDeepSeek4.67/5.0$0.210$0.790164K
42DeepSeek V4 ProDDeepSeek4.67/5.0$0.435$0.8701.0M
43Qwen: Qwen3.6 FlashQQwen4.67/5.0$0.188$1.131M
44GPT-5.4 NanoOOpenAI4.67/5.0$0.200$1.25400K
45Gemini 3.1 Flash Lite PreviewGGoogle4.67/5.0$0.250$1.501.0M
46Mistral Medium 3.1MMistral4.67/5.0$0.400$2.00131K
47R1 0528DDeepSeek4.67/5.0$0.500$2.15164K
48R1DDeepSeek4.67/5.0$0.700$2.50164K
49Grok 4.3XXAI4.67/5.0$1.25$2.501M
50Xiaomi: MiMo-V2-ProXXiaomi4.67/5.0$1.00$3.001.0M
51Mistral Medium 3.5MMistral4.67/5.0$1.50$7.50262K
52o3OOpenAI4.67/5.0$2.00$8.00200K
53Gemini 3.5 FlashGGoogle4.67/5.0$1.50$9.001.0M
54Gemini 2.5 ProGGoogle4.67/5.0$1.25$10.001.0M
55GPT-5.5OOpenAI4.67/5.0$5.00$30.001.1M
56OpenAI: gpt-oss-20bOOpenAI4.33/5.0$0.030$0.140131K
57Qwen: Qwen3.5-9BQQwen4.33/5.0$0.040$0.150262K
58Qwen: Qwen3 30B A3B Instruct 2507QQwen4.33/5.0$0.090$0.300262K
59GPT-5 NanoOOpenAI4.33/5.0$0.050$0.400400K
60Gemini 2.5 Flash LiteGGoogle4.33/5.0$0.100$0.4001.0M
61Grok 3 MiniXXAI4.33/5.0$0.300$0.500131K
62DeepSeek V3.1 TerminusDDeepSeek4.33/5.0$0.270$0.950164K
63Google: Gemini 3.1 Flash LiteGGoogle4.33/5.0$0.250$1.501.0M
64Mistral Large 3 2512MMistral4.33/5.0$0.500$1.50262K
65GPT-4.1 MiniOOpenAI4.33/5.0$0.400$1.601.0M
66Devstral 2 2512MMistral4.33/5.0$0.400$2.00262K
67Ministral 3 14B 2512MMistral4.00/5.0$0.200$0.200262K
68Llama 3.3 70B InstructMMeta4.00/5.0$0.100$0.320131K
69Mistral Small 4MMistral4.00/5.0$0.150$0.600262K
70Mistral Small 3.1 24BMMistral4.00/5.0$0.351$0.555128K
71Codestral 2508MMistral4.00/5.0$0.300$0.900256K
72Gemini 2.5 FlashGGoogle4.00/5.0$0.300$2.501.0M
73Ministral 3 3B 2512MMistral3.67/5.0$0.100$0.100131K
74Ministral 3 8B 2512MMistral3.67/5.0$0.150$0.150262K
75Qwen: Qwen3 Coder 30B A3B InstructQQwen3.67/5.0$0.070$0.270160K
76Llama 4 ScoutMMeta3.67/5.0$0.080$0.30010M
77GPT-4.1 NanoOOpenAI3.67/5.0$0.100$0.4001.0M
78Grok Code Fast 1XXAI3.67/5.0$0.200$1.50256K
79Mistral Small 3.2 24BMMistral3.33/5.0$0.075$0.200128K
80Devstral Small 1.1MMistral3.33/5.0$0.100$0.300131K
81Llama 4 MaverickMMeta3.33/5.0$0.150$0.6001.0M
82Devstral MediumMMistral3.33/5.0$0.400$2.00131K
83GPT-4oOOpenAI3.33/5.0$2.50$10.00128K
84GPT-4o-miniOOpenAI3.00/5.0$0.150$0.600128K
85Xiaomi: MiMo-V2-OmniXXiaomi3.00/5.0$0.400$2.00262K

Pricing — top 5 for research

QQwen: Qwen3 235B A22B Instruct 2507
$0.093/MTok
5.00/5.0
OOpenAI: gpt-oss-120b
$0.145/MTok
5.00/5.0
NNVIDIA: Nemotron 3 Nano 30B A3B
$0.163/MTok
5.00/5.0
XXiaomi: MiMo-V2-Flash
$0.250/MTok
5.00/5.0
GGemma 4 26B A4B
$0.263/MTok
5.00/5.0
modelpicker.aipowered by live benchmark data

The best AI for research changes every month.

We'll email you when rankings shift, new models hit the top 5, or pricing cuts reshuffle the value leaders.

Get notified when models change
Price drops, new models, benchmark updates. One email per change, no spam.