Skip to content

Drop in a screenshot and convert it to clean code (HTML/Tailwind/React/Vue)

Python

Refresh text-create model mix from judged text evals

Abi Rajab0d6ab882e61
diff --git a/backend/routes/model_choice_sets.py b/backend/routes/model_choice_sets.py
index 45cf3c0d..a3398e69 100644
--- a/backend/routes/model_choice_sets.py
+++ b/backend/routes/model_choice_sets.py
@@ -23,10 +23,14 @@ ALL_KEYS_MODELS_DEFAULT = (
# Text (Create)
+# Refreshed 2026-07-27 from the "Text evals v1" session (8 briefs x 7
+# models): sol high and opus 5 high replace gpt-5.5 high and opus 4.8 high
+# at similar cost/latency; 3-flash minimal keeps the fast slot (48s median,
+# $0.05/brief) and 3.1-pro low stays for now.
ALL_KEYS_MODELS_TEXT_CREATE = (
Llm.GEMINI_3_FLASH_PREVIEW_MINIMAL,
- Llm.GPT_5_5_HIGH,
- Llm.CLAUDE_OPUS_4_8_HIGH,
+ Llm.GPT_5_6_SOL_HIGH,
+ Llm.CLAUDE_OPUS_5_HIGH,
Llm.GEMINI_3_1_PRO_PREVIEW_LOW,
)
diff --git a/backend/tests/test_model_selection.py b/backend/tests/test_model_selection.py
index 321ac50a..24087dfc 100644
--- a/backend/tests/test_model_selection.py
+++ b/backend/tests/test_model_selection.py
@@ -25,8 +25,8 @@ class TestModelSelectionAllKeys:
expected = [
Llm.GEMINI_3_FLASH_PREVIEW_MINIMAL,
- Llm.GPT_5_5_HIGH,
- Llm.CLAUDE_OPUS_4_8_HIGH,
+ Llm.GPT_5_6_SOL_HIGH,
+ Llm.CLAUDE_OPUS_5_HIGH,
Llm.GEMINI_3_1_PRO_PREVIEW_LOW,
]
assert models == expected