Text2SQL on BIRD (dev)
65.8Exec Acc (Greedy)TRUST-SQL-8B
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TRUST-SQL-8BSchema Prefill=false2026.03 | 65.8 | 67.7 | — | |
| Reasoning-SQL-14BBase Model=Qwen2.5-Coder-14B-Instruct, Training Set=BIRD-train(8k)2026.01 | 65.3 | — | 0.34 | |
| TRUST-SQL-4BSchema Prefill=false2026.03 | 64.9 | 67.2 | — | |
| Omini-SQL-14BBase Model=Qwen2.5-Coder-14B-Instruct, Training Set=SynSQL(2.5M)+BIRD(9.4k)+Spider(8.7k)2026.01 | 64.2 | 65.9 | 0.001 | |
| Reasoning-SQL-7BBase Model=Qwen2.5-Coder-7B-Instruct, Training Set=BIRD-train(8026)2026.01 | 64 | — | 1.63 | |
| OminiSQL-7BBase Model=Qwen2.5-Coder-7B-Instruct, Training Set=SynSQL(2.5M)+BIRD(9.4k)+Spider(8.7k)2026.01 | 63.9 | 66.1 | 0.005 | |
| SynSQL(2.5M)Backbone=Qwen2.5-Coder-7B-Instruct2025.12 | 63.9 | 66.1 | — | |
| OmniSQL-7BSchema Prefill=true2026.03 | 63.9 | 66.1 | — | |
| SQL-R1-7BSchema Prefill=true2026.03 | 63.7 | 66.6 | — | |
| SQL-TRAIL-14BBase Model=Qwen2.5-Coder-14B-Instruct, Training Set=SynSQL(0.8k)+Spider(1k)2026.01 | 63.6 | 66.7 | 1.05 | |
| SQL-R1-14BBase Model=Qwen2.5-Coder-14B-Instruct, Training Set=SynSQL(5k)2026.01 | 63.2 | 67.1 | 0.47 | |
| MTIR-SQL-4BSchema Prefill=true2026.03 | 63.1 | 64.4 | — | |
| GPT-4-TurboAccess=Closed-source2025.12 | 62 | 63.6 | — | |
| GPT-4oAccess=Closed-source2025.12 | 61.9 | 64 | — | |
| Multi-turn AgentBase Model=Qwen2.5-Coder-14B-Instruct, Training Set=-2026.01 | 61.5 | 66.1 | — | |
| SQL-TRAIL-7BBase Model=Qwen2.5-Coder-7B-Instruct, Training Set=SynSQL(0.8k)+Spider(1k)2026.01 | 60.1 | 64.2 | 4.6 | |
| SQL-Trail-7BSchema Prefill=true2026.03 | 60.1 | 64.2 | — | |
| Multi-turn AgentBase Model=Sonnet-3.7, Training Set=-2026.01 | 60 | 60.8 | — | |
| DataFlow-Text2SQL-90KBackbone=Qwen2.5-Coder-7B-Instruct2025.12 | 59.2 | 61.5 | — | |
| SQL-R1-7BBase Model=Qwen2.5-Coder-7B-Instruct, Training Set=SynSQL(5k)2026.01 | 58.9 | 63.1 | 1.6 | |
| SynSQL(2.5M)Backbone=Meta-Llama-3.1-8B-Instruct2025.12 | 58.9 | 62 | — | |
| GPT-4o-miniAccess=Closed-source2025.12 | 58.8 | 61.5 | — | |
| Single-passBase Model=Sonnet-3.7, Training Set=-2026.01 | 58.5 | 60.1 | — | |
| Spider+BIRD+DataFlow-Text2SQL-90KBackbone=Qwen2.5-Coder-7B-Instruct2025.12 | 58.3 | 64 | — | |
| DataFlow-Text2SQL-50KBackbone=Qwen2.5-Coder-7B-Instruct2025.12 | 57.9 | 62.5 | — | |
| SynSQL(90K)Backbone=Qwen2.5-Coder-7B-Instruct2025.12 | 56.2 | 59.4 | — | |
| DataFlow-Text2SQL-90KBackbone=Meta-Llama-3.1-8B-Instruct2025.12 | 54.6 | 56.8 | — | |
| SynSQL(50K)Backbone=Qwen2.5-Coder-7B-Instruct2025.12 | 54 | 59.3 | — | |
| Spider+BIRD+DataFlow-Text2SQL-90KBackbone=Meta-Llama-3.1-8B-Instruct2025.12 | 53.4 | 58.9 | — | |
| DataFlow-Text2SQL-50KBackbone=Meta-Llama-3.1-8B-Instruct2025.12 | 51.4 | 57.6 | — | |
| SynSQL(90K)Backbone=Meta-Llama-3.1-8B-Instruct2025.12 | 51.1 | 54.9 | — | |
| Multi-turn AgentBase Model=Qwen2.5-Coder-7B-Instruct, Training Set=-2026.01 | 50.9 | 61.3 | — | |
| Qwen2.5-Coder-7B-InstructAccess=Open-source2025.12 | 50.9 | 61.3 | — | |
| SQL-TRAIL-3BBase Model=Qwen2.5-Coder-3B-Instruct, Training Set=SynSQL(0.8k)+Spider(1k)2026.01 | 50.1 | 55.1 | 2.4 | |
| SQL-Trail-3BSchema Prefill=true2026.03 | 50.1 | 55.1 | — | |
| SynSQL(50K)Backbone=Meta-Llama-3.1-8B-Instruct2025.12 | 49.1 | 55.2 | — | |
| SQL-R1-3BBase Model=Qwen2.5-Coder-3B-Instruct, Training Set=SynSQL(5k)2026.01 | 48.4 | 54.6 | 0.64 | |
| Qwen2.5-7B-InstructAccess=Open-source2025.12 | 46.9 | 56.4 | — | |
| Multi-turn AgentBase Model=Qwen2.5-Coder-3B-Instruct, Training Set=-2026.01 | 45.2 | 50.5 | — | |
| DeepSeek-Coder-7B-InstructAccess=Open-source2025.12 | 43.1 | 48 | — | |
| Meta-Llama-3.1-8B-InstructAccess=Open-source2025.12 | 42 | 53.1 | — | |
| OpenCoder-8B-InstructAccess=Open-source2025.12 | 37.5 | 45.3 | — | |
| Granite-3.1-8B-InstructAccess=Open-source2025.12 | 36 | 47.2 | — | |
| Granite-8B-Code-InstructAccess=Open-source2025.12 | 27.6 | 32.5 | — | |
| MTIR-SQL-8BSchema Prefill=true2026.03 | — | 64.6 | — | |
| SQL-R1-3BSchema Prefill=true2026.03 | — | 54.6 | — |