List your company
ResearchFormerly WizWand

Scientific Question Answering on GPQA Diamond

84.4AccuracyGemini 2.5 Pro
24.49640.04855.671.152Jul 14, 2025Sep 6, 2025Oct 31, 2025Dec 25, 2025Feb 18, 2026Apr 14, 2026Jun 8, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Gemini 2.5 Pro2026.0284.4
Interfaze-Beta2026.0281.31
Claude Opus 4Reasoning=Thinking2026.0279.6
Claude Sonnet 4Reasoning=Thinking2026.0277.7
Claude Sonnet 42026.0268.3
Gemini 2.5 Flash2026.0268.3
GPT-5Reasoning=Minimal Reasoning2026.0267.3
GPT-4.1(2025-04-14)Maximum output tokens=32,7682025.0767.17
GPT-o3-mini(2025-01-31)Maximum output tokens=32,7682025.0766.67
GPT-4.12026.0266.3
SMCSMaximum output tokens=32,7682025.0766.16
Qwen3-32BMaximum output tokens=32,7682025.0764.65
Claude-3.7-Sonnet(2025-02-19)Maximum output tokens=32,7682025.0763.64
EXAONE-Deep-32BMaximum output tokens=32,7682025.0763.13
QwQ-32BMaximum output tokens=32,7682025.0762.63
GLM-Z1-32B-0414Maximum output tokens=32,7682025.0762.12
Claude-3.5-Sonnet(2024-06-20)Maximum output tokens=32,7682025.0761.62
DeepSeek-R1-Distill-Llama-70BMaximum output tokens=32,7682025.0760.6
DeepSeek-V3Starting Checkpoint=Base, Reasoning Strategy=Self-Contained Reasoning, Backbone Architecture Family=DeepSeek-V32026.0259.1
ELPO-4BStarting Checkpoint=Inst, Backbone Architecture Family=Qwen3-4B2026.0259.1
Qwen-3-30B-A3BOpenness=Open-weights, Regional Origin=Non-European, Post-training=Instruction-tuned2026.0258.6
DemyAgent-4BStarting Checkpoint=Inst, Backbone Architecture Family=Qwen3-4B2026.0258.5
DeepSeek-R1-Distill-Qwen-32BMaximum output tokens=32,7682025.0757.58
Qwen-3-14BOpenness=Open-weights, Regional Origin=Non-European, Post-training=Instruction-tuned2026.0256.6
Qwen-3-32BOpenness=Open-weights, Regional Origin=Non-European, Post-training=Instruction-tuned2026.0254.7
Base(instruct)Architecture=Qwen3-30B-A3B2026.0554.04
ARPOStarting Checkpoint=Inst, Backbone Architecture Family=Qwen2.5-7B2026.0253
AEPOStarting Checkpoint=Inst, Backbone Architecture Family=Qwen2.5-7B2026.0253
ELPO-7BStarting Checkpoint=Inst, Backbone Architecture Family=Qwen2.5-7B2026.0252.7
GPT-4o(2024-08-06)Maximum output tokens=32,7682025.0752.53
HuatuoGPT-o1-72BMaximum output tokens=32,7682025.0752.53
GIGPOStarting Checkpoint=Inst, Backbone Architecture Family=Qwen2.5-7B2026.0252.5
GRPOArchitecture=Qwen3-30B-A3B2026.0552.02
Qwen3-4BStarting Checkpoint=Inst, Reasoning Strategy=Self-Contained Reasoning, Backbone Architecture Family=Qwen3-4B2026.0252
DemyAgentStarting Checkpoint=Inst, Backbone Architecture Family=Qwen2.5-7B2026.0252
ToRLStarting Checkpoint=Math-Inst, Backbone Architecture Family=Qwen2.5-7B2026.0251.5
Gemma-3-27b-itMaximum output tokens=32,7682025.0750.51
CIRStarting Checkpoint=Math, Backbone Architecture Family=Qwen2.5-7B2026.0249.5
MESAArchitecture=Qwen3-30B-A3B2026.0549.49
Stair-SFTArchitecture=Qwen3-30B-A3B2026.0548.98
Llama-3.3-Nemotron-Super-49B-v1Maximum output tokens=32,7682025.0748.48
OPDTeacher-Student Setting=DeepSeek-R1 -> DeepSeek-R1-Distill-Qwen-7B2026.0648.1
TIPSetting=Qwen3-14B -> Qwen3-4B, Supervised-token budget=10%2026.0547.98
BaseSetting=Qwen3-8B (GRPO) -> Qwen3-4B, Supervised-token budget=10%2026.0547.97
BaseSetting=Qwen3-14B -> Qwen3-4B, Supervised-token budget=10%2026.0547.97
EntropyModel=Qwen3-14B-Base, Evaluation Samples=8 samples2026.0147.85
Gemma-3-27BOpenness=Open-weights, Regional Origin=Non-European, Post-training=Instruction-tuned2026.0247.6
ALMTeacher-Student Setting=DeepSeek-R1 -> DeepSeek-R1-Distill-Qwen-7B2026.0647.6
Mistral-3.2-24BOpenness=Open-weights, Regional Origin=European, Post-training=Instruction-tuned2026.0247.5
EntropySetting=Qwen3-8B (GRPO) -> Qwen3-4B, Supervised-token budget=10%2026.0547.47
TIPSetting=Qwen3-8B (GRPO) -> Qwen3-4B, Supervised-token budget=10%2026.0547.47
TA-OPDSetting=Qwen3-8B (GRPO) -> Qwen3-4B, Supervised-token budget=10%2026.0547.47
Stair-DPOArchitecture=Qwen3-30B-A3B2026.0547.47
CDMTeacher-Student Setting=DeepSeek-R1 -> DeepSeek-R1-Distill-Qwen-7B2026.0647.4
Pure OPDSetting=Qwen3-8B (GRPO) -> Qwen3-4B, Supervised-token budget=10%2026.0547.22
Qwen-2.5-72B-InstructMaximum output tokens=32,7682025.0746.97
Llama-3.3-70B-InstructMaximum output tokens=32,7682025.0746.97
SFTTeacher-Student Setting=DeepSeek-R1 -> DeepSeek-R1-Distill-Qwen-7B2026.0646.9
Llama-3.3-70BOpenness=Open-weights, Regional Origin=Non-European, Post-training=Instruction-tuned2026.0246.6
ProFitModel=Qwen3-14B-Base, Evaluation Samples=8 samples2026.0146.53
EntropySetting=Qwen3-14B -> Qwen3-4B, Supervised-token budget=10%2026.0546.46
SFTModel=Qwen3-14B-Base, Evaluation Samples=8 samples2026.0146.02
TA-OPD+Ent.Setting=Qwen3-8B (GRPO) -> Qwen3-4B, Supervised-token budget=10%2026.0545.7
TA-OPDSetting=Qwen3-14B -> Qwen3-4B, Supervised-token budget=10%2026.0545.2
SafeXArchitecture=Qwen3-30B-A3B2026.0544.44
Qwen3-4BStarting Checkpoint=Inst, Reasoning Strategy=TIR Reasoning, Backbone Architecture Family=Qwen3-4B2026.0244.3
DFTModel=Qwen3-14B-Base, Evaluation Samples=8 samples2026.0143.81
Pure OPDSetting=Qwen3-14B -> Qwen3-4B, Supervised-token budget=10%2026.0542.17
TA-OPD+Ent.Setting=Qwen3-14B -> Qwen3-4B, Supervised-token budget=10%2026.0542.17
Qwen2.5-Coder-32B-InstructMaximum output tokens=32,7682025.0741.92
Qwen2.5-32b-InstructMaximum output tokens=32,7682025.0740.91
SFTArchitecture=Qwen3-30B-A3B2026.0540.4
VanillaModel=Qwen3-14B-Base, Evaluation Samples=8 samples2026.0137.69
Gemma-3-12BOpenness=Open-weights, Regional Origin=Non-European, Post-training=Instruction-tuned2026.0237.2
OLMo-3.1-32BOpenness=Fully-open, Regional Origin=Non-European, Post-training=Instruction-tuned2026.0236
EntropyModel=Qwen3-4B-Base, Evaluation Samples=8 samples2026.0134.91
InternLM2.5-20B-ChatMaximum output tokens=32,7682025.0734.85
TA-OPDSetting=Qwen3-4B -> Qwen3-1.7B, Supervised-token budget=10%2026.0534.59
ProFitModel=Qwen3-4B-Base, Evaluation Samples=8 samples2026.0134.34
SFTModel=Qwen3-4B-Base, Evaluation Samples=8 samples2026.0134.15
DAPOStarting Checkpoint=Instruct, Backbone Architecture Family=Qwen2.5-7B2026.0233.5
TeleChat2-35B-32KMaximum output tokens=32,7682025.0733.33
OLMo-3-7BOpenness=Fully-open, Regional Origin=Non-European, Post-training=Instruction-tuned2026.0233.2
GRPOStarting Checkpoint=Inst, Backbone Architecture Family=Qwen2.5-7B2026.0232.7
EntropySetting=Qwen3-4B -> Qwen3-1.7B, Supervised-token budget=10%2026.0532.58
TA-OPD+Ent.Setting=Qwen3-4B -> Qwen3-1.7B, Supervised-token budget=10%2026.0532.58
DFTModel=Qwen3-4B-Base, Evaluation Samples=8 samples2026.0131.69
TA-OPD+Ent.Setting=DeepSeek-R1-Distill-Qwen-14B -> Qwen2.5-3B, Supervised-token budget=10%2026.0531.63
BaseSetting=Qwen3-4B -> Qwen3-1.7B, Supervised-token budget=10%2026.0531.56
Qwen2.5-7BStarting Checkpoint=Inst, Reasoning Strategy=Self-Contained Reasoning, Backbone Architecture Family=Qwen2.5-7B2026.0231.3
Reinforce++Starting Checkpoint=Instruct, Backbone Architecture Family=Qwen2.5-7B2026.0231.3
SFTArchitecture=DeepSeek-v2-Lite2026.0530.81
Pure OPDSetting=Qwen3-4B -> Qwen3-1.7B, Supervised-token budget=10%2026.0530.55
TA-OPDSetting=DeepSeek-R1-Distill-Qwen-14B -> Qwen2.5-3B, Supervised-token budget=10%2026.0529.59
TIPSetting=Qwen3-4B -> Qwen3-1.7B, Supervised-token budget=10%2026.0529.29
Stair-SFTArchitecture=DeepSeek-v2-Lite2026.0528.28
BaseSetting=DeepSeek-R1-Distill-Qwen-14B -> Qwen2.5-3B, Supervised-token budget=10%2026.0527.81
Pure OPDSetting=DeepSeek-R1-Distill-Qwen-14B -> Qwen2.5-3B, Supervised-token budget=10%2026.0527.78
TIPSetting=DeepSeek-R1-Distill-Qwen-14B -> Qwen2.5-3B, Supervised-token budget=10%2026.0527.55
EuroLLM-22BOpenness=Fully-open, Regional Origin=European, Post-training=Instruction-tuned2026.0226.8