feat: add gpt-6.1-sol to the chat, agent steps and weekly ai evals (#11584)

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Ruben Fiszel <ruben@windmill.dev>

.github/workflows/ai-evals-scheduled.yml

@@ -15,8 +15,8 @@ on:
         description: "Space-separated modes (global, cli, flow, app, script)"
         default: "global cli"
       models:
-        description: "Space-separated model aliases (bun run cli -- models); cli skips those it has no agent for"
-        default: "sonnet-5.5 opus-5.5 gpt-6-sol gpt-6-astra deepseek-v4-pro"
+        description: "Space-separated model aliases (bun run cli -- models); cli skips those it has no agent for""
+        default: "sonnet-5.5 opus-5.5 gpt-6.1-sol gpt-6-astra deepseek-v4-pro"
       runs:
         description: "Runs per case"
         default: "3"
@@ -28,7 +28,7 @@ concurrency:
   group: ai-evals-scheduled-${{ github.ref }}
 
 env:
-  MODELS: ${{ inputs.models || 'sonnet-5.5 opus-5.5 gpt-6-sol gpt-6-astra deepseek-v4-pro' }}
+  MODELS: ${{ inputs.models || 'sonnet-5.5 opus-5.5 gpt-6.1-sol gpt-6-astra deepseek-v4-pro' }}
   RUNS: ${{ inputs.runs || '3' }}
   REASONING: ${{ inputs.reasoning }}
   # Routes on the cloud instance are prefixed with their workspace.

ai_evals/core/models.ts

@@ -148,6 +148,15 @@ export const EVAL_MODELS: EvalModelSpec[] = [
       model: "gpt-6-astra",
     },
   },
+  {
+    id: "gpt-6.1-sol",
+    label: "GPT-6.1 Sol",
+    aliases: ["gpt-6.1-sol"],
+    frontend: {
+      provider: "openai",
+      model: "gpt-6.1-sol",
+    },
+  },
   {
     id: "gpt-6-sol",
     label: "GPT-6 Sol",

backend/windmill-ai/src/providers/mod.rs

@@ -62,9 +62,13 @@ const REASONING_RULES: &[ReasoningRule] = &[
         can_disable: false,
         completions_tools_need_off: false,
     },
-    // Live-verified: astra takes low..max only, where sol and luna also take `none`.
+    // Live-verified: gpt-6-astra and gpt-6.1-sol take low..max only, where gpt-6-sol and
+    // gpt-6-luna also take `none`.
     ReasoningRule {
-        matches: |m| base_id(m).starts_with("gpt-6-astra"),
+        matches: |m| {
+            let base = base_id(m);
+            base.starts_with("gpt-6-astra") || base.starts_with("gpt-6.1-sol")
+        },
         can_disable: false,
         completions_tools_need_off: true,
     },

frontend/src/lib/components/copilot/lib.ts

@@ -71,6 +71,7 @@ interface AIProviderDetails {
 // the frontier model. The gpt-5 family is deprecated (retires 2026-12-11) but
 // still served, so it stays in the list below the 5.6 models.
 const OPENAI_MODELS = [
+	'gpt-6.1-sol',
 	'gpt-6-sol',
 	'gpt-6-astra',
 	'gpt-6-luna',

frontend/src/lib/components/copilot/reasoningParity.json

@@ -47,6 +47,12 @@
 		"canDisable": false,
 		"completionsToolsNeedOff": true
 	},
+	{
+		"provider": "openai",
+		"model": "gpt-6.1-sol",
+		"canDisable": false,
+		"completionsToolsNeedOff": true
+	},
 	{
 		"provider": "openai",
 		"model": "gpt-6-sol",

frontend/src/lib/components/copilot/reasoningRegistry.ts

@@ -131,9 +131,10 @@ const BEDROCK_RULES: ReasoningRule[] = [
 // Azure names gpt-3.5 `gpt-35-turbo`. `minimal` exists on gpt-5 only, `xhigh` from
 // gpt-5.5, `max` from gpt-5.6.
 const OPENAI_RULES: ReasoningRule[] = [
-	// Live-verified: astra takes low..max only, where sol and luna also take `none`.
+	// Live-verified: gpt-6-astra and gpt-6.1-sol take low..max only, where gpt-6-sol and
+	// gpt-6-luna also take `none`.
 	{
-		match: /(?:^|\/)gpt-6-astra/,
+		match: /(?:^|\/)gpt-6(?:-astra|\.1-sol)/,
 		levels: LOW_TO_MAX,
 		canDisable: false,
 		completionsToolsNeedOff: true