Score change log
Every change to a published score, with the date and the reason. Scores never change silently.
In short
Newly Stack publishes every change to a benchmark score here: when it was published, corrected after a dispute or re-test, or withdrawn, always with a written reason. A database rule blocks any score change that has no log entry.
| Date | Tool or model | Suite | Change | Reason |
|---|---|---|---|---|
| Recraft | test-lab photo editing (ID) | 0 | Recraft photo-editing result withdrawn: the API route we used (OpenRouter recraft/recraft-v4) has no edit mode, so the model repainted the photo instead of editing it. That says nothing about Recraft's own editor, so the 0 of 5 result is no longer shown as a score. Raw outputs are kept. | |
| Command A+ | ns-id-v1 | 87.83 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Mistral Small 4 | ns-id-v1 | 84.5 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Mistral Medium 3.5 | ns-id-v1 | 89.6 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Qwen3.8 Max (0902) | ns-id-v1 | 97.55 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Kimi K3.1 Pro | ns-id-v1 | 99.26 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| DeepSeek V4.1 Flash | ns-id-v1 | 97.13 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| DeepSeek V4 Pro | ns-id-v1 | 96.8 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Grok 4.7 | ns-id-v1 | 97.98 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Perplexity Sonar | ns-id-v1 | 97.11 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Llama 4 Maverick | ns-id-v1 | 88.15 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Claude Haiku 5.5 | ns-id-v1 | 93.19 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Claude Sonnet 5.5 | ns-id-v1 | 97.64 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Claude Opus 5.5 | ns-id-v1 | 98.97 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Gemini 3.5 Flash-Lite | ns-id-v1 | 90.69 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Gemini 3.8 Flash | ns-id-v1 | 93 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| Gemini 4 Argon | ns-id-v1 | 94.58 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| GPT-6 Luna | ns-id-v1 | 97.4 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| GPT-6.1 Sol | ns-id-v1 | 98.46 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| GPT-6 Astra | ns-id-v1 | 97.99 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. | |
| ChatGPT (chat-latest) | ns-id-v1 | 96.4 | First publication of NS-ID v1 (run of 25 October 2026, model-graded), reviewed before launch. |
Answered score disputes
No score disputes have been answered yet.
Are you the maker of a tested tool? You can dispute a score within 30 days. Dispute a score