VGT AI Benchmark / Observatory
Benchmark explorer
Search and inspect published benchmark evidence with functional server-rendered deep links.
Claude Fable 5.1Anthropic · SELF_REPORTEDClaude Mythos 5.1Anthropic · SELF_REPORTEDClaude Opus 5.5Anthropic · SELF_REPORTEDClaude Sonnet 5.5Anthropic · SELF_REPORTEDCodestral 26.09Mistral AI · SELF_REPORTEDCommand A ReasoningCohere · SELF_REPORTEDDeepSeek R2DeepSeek · SELF_REPORTEDDeepSeek V4DeepSeek · SELF_REPORTEDDeepSeek V4.1 FlashDeepSeek · SELF_REPORTEDGLM-5.3-FlashZ.ai · SELF_REPORTEDGPT-6 AstraOpenAI · SELF_REPORTEDGPT-6 LunaOpenAI · SELF_REPORTEDGPT-6 SolOpenAI · SELF_REPORTEDGPT-Image 2.5 FlareOpenAI · SELF_REPORTEDGPT-Image 2.5 SunburstOpenAI · SELF_REPORTEDGPT-Live 1OpenAI · SELF_REPORTEDGemini 3.8 FlashGoogle · VERIFIEDGemini 3.8 Flash TTSGoogle · SELF_REPORTEDGemini 3.8 Flash-Lite TTSGoogle · SELF_REPORTEDGemini 3.8 LiveGoogle · SELF_REPORTEDGemini 3.8 Live Extended ThinkingGoogle · SELF_REPORTEDGemini 3.8 ProGoogle · SELF_REPORTEDGrok 4.7xAI · SELF_REPORTEDGrok Voice Transcribe 2.0xAI · SELF_REPORTEDKimi k2.5 ThinkingMoonshot AI · SELF_REPORTEDLlama 4 MaverickMeta · SELF_REPORTEDLlama 4 ScoutMeta · SELF_REPORTEDLyria 3.5Google · SELF_REPORTEDMistral Large 3Mistral AI · SELF_REPORTEDNorth Small Translate 1.0Cohere · SELF_REPORTEDQwen 3.5 Coder 32BAlibaba Cloud · SELF_REPORTEDQwen 3.5 MaxAlibaba Cloud · SELF_REPORTED
VGT AI Benchmark / Observatory
Model profile
GPT-Image 2.5 Flare by OpenAI
OPENAI
GPT-Image 2.5 Flare
2.5-flare · api
- Verification
- SELF_REPORTED
- Context
- 32.768
- Parameters
- —
Capability profile
Latest completed category scores.
No completed category scores yet.
Declared capabilities
text image_input image_output
No public data.No verified record currently matches this view.
Run history
Benchmark version, cost evidence and latency remain attached to each run rather than averaged into a vague model label.
No public data.No verified record currently matches this view.
Efficiency relationships
Quality is shown against measured cost and latency. Lower x-values are more efficient; the accessible table carries the same evidence.
Score vs cost
Score vs latency
No public data.No verified record currently matches this view.
Calibration
Confidence should track observed correctness. A diagonal indicates ideal calibration.
No public data.No verified record currently matches this view.
VGT AI Benchmark / Observatory
Explorer leaderboard
Hard benchmark evidence with explicit version and verification context.
1
Gemini 3.8 FlashGoogle · 3.8-flash
100,00 %
2
Gemini 3.8 FlashGoogle · 3.8-flash
100,00 %
3
Gemini 3.8 FlashGoogle · 3.8-flash
100,00 %
| Model | Provider | Category | Score | Verification | Benchmark |
|---|---|---|---|---|---|
| Gemini 3.8 Flash | coding | 100,00 % | VERIFIED | BEN_vgt_core_suite_2026_v1 1.0.0 | |
| Gemini 3.8 Flash | mathematics | 100,00 % | VERIFIED | BEN_vgt_core_suite_2026_v1 1.0.0 | |
| Gemini 3.8 Flash | reasoning | 100,00 % | VERIFIED | BEN_vgt_core_suite_2026_v1 1.0.0 |