claude-evolve 1.11.18 → 1.11.19

package/bin/claude-evolve-check

@@ -90,7 +90,7 @@ SCRIPT
     gemini-pro)
       cat > "$test_script" << 'SCRIPT'
 #!/usr/bin/env bash
-exec gemini -y -m gemini-3-pro-preview -p "$1"
+exec gemini -y -m auto-gemini-3 -p "$1"
 SCRIPT
       ;;
     gemini-flash|gemini-3-flash)
@@ -141,10 +141,22 @@ SCRIPT
 exec codex exec --dangerously-bypass-approvals-and-sandbox "$1"
 SCRIPT
       ;;
-    gpt-5-codex)
+    codex-think)
       cat > "$test_script" << 'SCRIPT'
 #!/usr/bin/env bash
-exec codex exec -m gpt-5-codex --dangerously-bypass-approvals-and-sandbox "$1"
+exec codex exec -m gpt-5.4 -c model_reasoning_effort="high" --dangerously-bypass-approvals-and-sandbox "$1"
+SCRIPT
+      ;;
+    codex-coding)
+      cat > "$test_script" << 'SCRIPT'
+#!/usr/bin/env bash
+exec codex exec -m gpt-5.4 -c model_reasoning_effort="medium" --dangerously-bypass-approvals-and-sandbox "$1"
+SCRIPT
+      ;;
+    gpt-5.4)
+      cat > "$test_script" << 'SCRIPT'
+#!/usr/bin/env bash
+exec codex exec -m gpt-5.4 --dangerously-bypass-approvals-and-sandbox "$1"
 SCRIPT
       ;;
     gpt-5.2)
@@ -159,22 +171,22 @@ SCRIPT
 exec codex exec -m gpt-5.3-codex --dangerously-bypass-approvals-and-sandbox "$1"
 SCRIPT
       ;;
-    gpt-5.3-codex-spark)
+    codex-spark|gpt-5.1-codex-mini)
       cat > "$test_script" << 'SCRIPT'
 #!/usr/bin/env bash
-exec codex exec -m gpt-5.3-codex-spark --dangerously-bypass-approvals-and-sandbox "$1"
+exec codex exec -m gpt-5.1-codex-mini --dangerously-bypass-approvals-and-sandbox "$1"
 SCRIPT
       ;;
     gemini-5-flash)
       cat > "$test_script" << 'SCRIPT'
 #!/usr/bin/env bash
-exec gemini -y -m gemini-5-flash -p "$1"
+exec gemini -y -m gemini-3-flash-preview -p "$1"
 SCRIPT
       ;;
     qwen-openrouter)
       cat > "$test_script" << 'SCRIPT'
 #!/usr/bin/env bash
-exec opencode -m openrouter/qwen/qwen3.5-plus-02-15 run "$1"
+exec opencode -m openrouter/qwen/qwen3.6-plus:free run "$1"
 SCRIPT
       ;;
     *)

package/lib/ai-cli.sh

@@ -105,10 +105,22 @@ $prompt"
       ai_output=$(codex exec -m "$codex_gpt5_model" --dangerously-bypass-approvals-and-sandbox "$prompt" 2>&1)
       local ai_exit_code=$?
       ;;
-    gpt-5-codex)
+    codex-think)
       local ai_output
-      # GPT-5 Codex - code-specialized variant via Codex CLI
-      ai_output=$(codex exec -m gpt-5-codex --dangerously-bypass-approvals-and-sandbox "$prompt" 2>&1)
+      # GPT-5.4 high reasoning - for ideation tasks requiring deep thinking
+      ai_output=$(codex exec -m gpt-5.4 -c model_reasoning_effort="high" --dangerously-bypass-approvals-and-sandbox "$prompt" 2>&1)
+      local ai_exit_code=$?
+      ;;
+    codex-coding)
+      local ai_output
+      # GPT-5.4 medium reasoning - for coding/implementation tasks
+      ai_output=$(codex exec -m gpt-5.4 -c model_reasoning_effort="medium" --dangerously-bypass-approvals-and-sandbox "$prompt" 2>&1)
+      local ai_exit_code=$?
+      ;;
+    gpt-5.4)
+      local ai_output
+      # GPT-5.4 - latest frontier agentic coding model via Codex CLI
+      ai_output=$(codex exec -m gpt-5.4 --dangerously-bypass-approvals-and-sandbox "$prompt" 2>&1)
       local ai_exit_code=$?
       ;;
     gpt-5.2)
@@ -123,10 +135,10 @@ $prompt"
       ai_output=$(codex exec -m gpt-5.3-codex --dangerously-bypass-approvals-and-sandbox "$prompt" 2>&1)
       local ai_exit_code=$?
       ;;
-    gpt-5.3-codex-spark)
+    codex-spark|gpt-5.1-codex-mini)
       local ai_output
-      # GPT-5.3 Codex Spark - lightweight fallback via Codex CLI
-      ai_output=$(codex exec -m gpt-5.3-codex-spark --dangerously-bypass-approvals-and-sandbox "$prompt" 2>&1)
+      # GPT-5.1 Codex Mini - cheap/fast lightweight fallback via Codex CLI
+      ai_output=$(codex exec -m gpt-5.1-codex-mini --dangerously-bypass-approvals-and-sandbox "$prompt" 2>&1)
       local ai_exit_code=$?
       ;;
     o3high)
@@ -136,8 +148,8 @@ $prompt"
       ;;
     gemini-pro)
       local ai_output
-      # Gemini streams output while working
-      ai_output=$(gemini -y -m gemini-3-pro-preview -p "$prompt" 2>&1)
+      # Gemini 3 auto-routing (gemini-3.1-pro / gemini-3-flash) - streams output while working
+      ai_output=$(gemini -y -m auto-gemini-3 -p "$prompt" 2>&1)
       local ai_exit_code=$?
       ;;
     gemini-flash)
@@ -149,7 +161,7 @@ $prompt"
     gemini-5-flash)
       local ai_output
       # Gemini 5 Flash - cheap fallback model
-      ai_output=$(gemini -y -m gemini-5-flash -p "$prompt" 2>&1)
+      ai_output=$(gemini -y -m gemini-3-flash-preview -p "$prompt" 2>&1)
       local ai_exit_code=$?
       ;;
     gemini-3-pro-preview)
@@ -254,9 +266,9 @@ $prompt"
       ;;
     qwen-openrouter)
       local ai_output
-      # Qwen latest - Alibaba's flagship model (currently qwen3.5-plus)
+      # Qwen latest - Alibaba's flagship model (currently qwen3.6-plus, free promotional tier)
       # Linear attention + sparse MoE, strong multimodal capabilities
-      ai_output=$(opencode -m openrouter/qwen/qwen3.5-plus-02-15 run "$prompt" 2>&1)
+      ai_output=$(opencode -m openrouter/qwen/qwen3.6-plus:free run "$prompt" 2>&1)
       local ai_exit_code=$?
       ;;
     codex-oss-local)

package/lib/ai_cli.py

@@ -248,8 +248,9 @@ MODEL_TIMEOUTS = {
     'gemini-pro': 1800, 'gemini-flash': 1200, 'gemini-3-flash': 600,
     'gemini-3-pro-preview': 1800, 'gemini-5-flash': 600,
     # Codex/OpenAI models - 10 min standard
-    'gpt-5-codex': 600, 'gpt-5.2': 600, 'gpt-5.3-codex': 600,
-    'gpt-5.3-codex-spark': 300,
+    'codex-think': 900, 'codex-coding': 600, 'codex-spark': 300,
+    'gpt-5.4': 600, 'gpt-5.2': 600, 'gpt-5.3-codex': 600,
+    'gpt-5.1-codex-mini': 300,
     # Z.AI agentic modes - 30 min (can be slow)
     'glm-zai': 1800, 'glm-5-zai': 1800,
     # Codex local - 40 min (local inference can be slow)

package/lib/config.sh

@@ -63,13 +63,13 @@ DEFAULT_WORKER_MAX_CANDIDATES=3
 #
 # Run: Subscription-based agentic models for code generation
 # All CLI tools (opencode, claude, gemini, kimi) are agentic and can edit files
-DEFAULT_LLM_RUN="gemini-pro gemini-pro glm-5-zai glm-5-zai kimi-coder kimi-coder gpt-5-codex gpt-5-codex sonnet"
-DEFAULT_LLM_RUN_FALLBACK="haiku glm-5-zai gemini-5-flash gpt-5.3-codex-spark"
+DEFAULT_LLM_RUN="gemini-pro gemini-pro glm-5-zai glm-5-zai kimi-coder kimi-coder codex-coding codex-coding sonnet"
+DEFAULT_LLM_RUN_FALLBACK="haiku glm-5-zai gemini-5-flash codex-spark"
 #
 # Ideate: Agentic models that can edit files for ideation
 # All CLI tools (opencode, claude, gemini, kimi) are agentic and can edit files
-DEFAULT_LLM_IDEATE="opus-think glm-5-zai gemini-pro kimi-coder gpt-5.2 gpt-5.3-codex qwen-openrouter"
-DEFAULT_LLM_IDEATE_FALLBACK="haiku glm-5-zai gemini-5-flash gpt-5.3-codex-spark"
+DEFAULT_LLM_IDEATE="opus-think glm-5-zai gemini-pro kimi-coder gpt-5.2 codex-think qwen-openrouter"
+DEFAULT_LLM_IDEATE_FALLBACK="haiku glm-5-zai gemini-5-flash codex-spark"
 
 # Load configuration from a YAML file and update variables
 _load_yaml_config() {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-evolve",
-  "version": "1.11.18",
+  "version": "1.11.19",
   "bin": {
     "claude-evolve": "bin/claude-evolve",
     "claude-evolve-main": "bin/claude-evolve-main",

package/templates/config.yaml

@@ -94,10 +94,10 @@ llm_cli:
 
   # Default configuration: sonnet at ~11%, rest doubled for cost savings
   # Commented out because these change over time; uncomment to override
-  #run: gemini-pro gemini-pro glm-5-zai glm-5-zai kimi-coder kimi-coder gpt-5-codex gpt-5-codex sonnet
-  #ideate: opus-think glm-5-zai gemini-pro kimi-coder gpt-5.2 gpt-5.3-codex
-  #run_fallback: haiku glm-5-zai gemini-5-flash gpt-5.3-codex-spark
-  #ideate_fallback: haiku glm-5-zai gemini-5-flash gpt-5.3-codex-spark
+  #run: gemini-pro gemini-pro glm-5-zai glm-5-zai kimi-coder kimi-coder codex-coding codex-coding sonnet
+  #ideate: opus-think glm-5-zai gemini-pro kimi-coder gpt-5.2 codex-think qwen-openrouter
+  #run_fallback: haiku glm-5-zai gemini-5-flash codex-spark
+  #ideate_fallback: haiku glm-5-zai gemini-5-flash codex-spark
 
   # Available models:
   # Claude (subscription-based, watch usage limits):
@@ -108,10 +108,12 @@ llm_cli:
   # - haiku: Claude Haiku via Claude CLI (cheap fallback)
   #
   # Codex/OpenAI (subscription-based):
-  # - gpt-5-codex: GPT-5 Codex (code-specialized) via Codex CLI
+  # - codex-think: GPT-5.4 high reasoning effort (ideation)
+  # - codex-coding: GPT-5.4 medium reasoning effort (coding/run)
+  # - codex-spark: GPT-5.1 Codex Mini (lightweight fallback)
+  # - gpt-5.4: GPT-5.4 no reasoning effort override via Codex CLI
   # - gpt-5.2: GPT-5.2 via Codex CLI
-  # - gpt-5.3-codex: GPT-5.3 Codex via Codex CLI
-  # - gpt-5.3-codex-spark: GPT-5.3 Codex Spark (lightweight fallback) via Codex CLI
+  # - gpt-5.3-codex: GPT-5.3 Codex (code-specialized) via Codex CLI
   # - gpt5: GPT-5 via Codex CLI (legacy alias)
   # - gpt5high: GPT-5 via Codex CLI (high reasoning)
   # - o3high: O3 via Codex CLI (high reasoning)