cecli-dev 0.97.4__tar.gz → 0.98.0__tar.gz

{cecli_dev-0.97.4 → cecli_dev-0.98.0}/CONTRIBUTING.md

@@ -159,4 +159,4 @@ The project's documentation is built using Jekyll and hosted on GitHub Pages. To
    bundle exec jekyll serve
    ```
 
-The built documentation will be available in the `aider/website/_site` directory.
+The built documentation will be available in the `cecli/website/_site` directory.

{cecli_dev-0.97.4 → cecli_dev-0.98.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cecli-dev
-Version: 0.97.4
+Version: 0.98.0
 Summary: we also can't pronounce cecli
 Project-URL: Homepage, https://github.com/dwash96/cecli
 Project-URL: Documentation, https://github.com/dwash96/cecli
@@ -35,7 +35,7 @@ Requires-Dist: beautifulsoup4>=4.13.4
 Requires-Dist: PyYAML>=6.0.2
 Requires-Dist: diff-match-patch>=20241021
 Requires-Dist: pypandoc>=1.15
-Requires-Dist: litellm>=1.80.11
+Requires-Dist: litellm!=1.82.7,!=1.82.8,>=1.80.11
 Requires-Dist: flake8>=7.3.0
 Requires-Dist: importlib_resources
 Requires-Dist: pyperclip>=1.9.0
@@ -74,6 +74,8 @@ Requires-Dist: cogapp; extra == "dev"
 Requires-Dist: semver; extra == "dev"
 Requires-Dist: codespell; extra == "dev"
 Requires-Dist: uv; extra == "dev"
+Requires-Dist: memray; extra == "dev"
+Requires-Dist: objgraph; extra == "dev"
 Provides-Extra: help
 Requires-Dist: llama-index-embeddings-huggingface; extra == "help"
 Requires-Dist: numpy>=1.26.4; extra == "help"
@@ -249,9 +251,9 @@ The current priorities are to improve core capabilities and user experience of t
 
 7. **Sub Agents**
   * [ ] Add `/fork` and `/rejoin` commands to manually manage parts of the conversation history
-  * [ ] Add an instance-able view of the conversation system so sub agents get their own context and workspaces
-  * [ ] Modify coder classes to have discrete identifiers for themselves/management utilities for them to have their own slices of the world
-  * [ ] Refactor global files like todo lists to live inside instance folders to avoid state conflicts
+  * [x] Add an instance-able view of the conversation system so sub agents get their own context and workspaces
+  * [x] Modify coder classes to have discrete identifiers for themselves/management utilities for them to have their own slices of the world
+  * [x] Refactor global files like todo lists to live inside instance folders to avoid state conflicts
   * [ ] Add a `spawn` tool that launches a sub agent as a background command that the parent model waits for to finish
   * [ ] Add visibility into active sub agent calls in TUI
 

{cecli_dev-0.97.4 → cecli_dev-0.98.0}/README.md

@@ -165,9 +165,9 @@ The current priorities are to improve core capabilities and user experience of t
 
 7. **Sub Agents**
   * [ ] Add `/fork` and `/rejoin` commands to manually manage parts of the conversation history
-  * [ ] Add an instance-able view of the conversation system so sub agents get their own context and workspaces
-  * [ ] Modify coder classes to have discrete identifiers for themselves/management utilities for them to have their own slices of the world
-  * [ ] Refactor global files like todo lists to live inside instance folders to avoid state conflicts
+  * [x] Add an instance-able view of the conversation system so sub agents get their own context and workspaces
+  * [x] Modify coder classes to have discrete identifiers for themselves/management utilities for them to have their own slices of the world
+  * [x] Refactor global files like todo lists to live inside instance folders to avoid state conflicts
   * [ ] Add a `spawn` tool that launches a sub agent as a background command that the parent model waits for to finish
   * [ ] Add visibility into active sub agent calls in TUI
 

{cecli_dev-0.97.4 → cecli_dev-0.98.0}/benchmark/benchmark.py

@@ -515,7 +515,8 @@ def load_results(results_dir, stats_languages=None):
                 results["test_cases_total"] = total_cases
             if passed_cases is not None:
                 results["test_cases_passed"] = passed_cases
-
+            if passed_cases is not None and total_cases is not None:
+                results["test_cases_pass_ratio"] = passed_cases / total_cases
             # Update the JSON file immediately to keep data fresh
             fname.write_text(json.dumps(results, indent=4))
             logger.debug(f"Updated {fname} with test case counts")
@@ -583,7 +584,8 @@ def summarize_results(results_dir, verbose, stats_languages=None):
     res.completion_tokens = 0
     res.test_cases_total = 0
     res.test_cases_passed = 0
-
+    res.test_cases_pass_ratio_sum = 0.0
+    res.test_cases_pass_ratio_count = 0
     res.reasoning_effort = None
     res.thinking_tokens = None
     res.map_tokens = None
@@ -621,6 +623,8 @@ def summarize_results(results_dir, verbose, stats_languages=None):
         lang_stats.completion_tokens = 0
         lang_stats.test_cases_total = 0
         lang_stats.test_cases_passed = 0
+        lang_stats.test_cases_pass_ratio_sum = 0.0
+        lang_stats.test_cases_pass_ratio_count = 0
         lang_to_stats[lang] = lang_stats
         lang_to_passed_tests[lang] = [0] * tries
 
@@ -682,6 +686,11 @@ def summarize_results(results_dir, verbose, stats_languages=None):
             if passed_cases is not None:
                 add("test_cases_passed", passed_cases, res, lang_stats)
 
+            pass_ratio = results.get("test_cases_pass_ratio")
+            if pass_ratio is not None:
+                add("test_cases_pass_ratio_sum", pass_ratio, res, lang_stats)
+                add("test_cases_pass_ratio_count", 1, res, lang_stats)
+
             res.reasoning_effort = results.get("reasoning_effort")
             res.thinking_tokens = results.get("thinking_tokens")
             res.map_tokens = results.get("map_tokens")
@@ -764,6 +773,12 @@ def summarize_results(results_dir, verbose, stats_languages=None):
         res.test_cases_percentage = 100 * res.test_cases_passed / res.test_cases_total
         print(f"  test_cases_percentage: {res.test_cases_percentage:.1f}")
 
+    if res.test_cases_pass_ratio_count > 0:
+        res.test_cases_pass_ratio_avg = (
+            100 * res.test_cases_pass_ratio_sum / res.test_cases_pass_ratio_count
+        )
+        print(f"  test_cases_pass_ratio_avg: {res.test_cases_pass_ratio_avg:.1f}")
+
     if variants["model"]:
         a_model = set(variants["model"]).pop()
         command = f"cecli --model {a_model}"
@@ -801,6 +816,18 @@ def summarize_results(results_dir, verbose, stats_languages=None):
                 pass_rate = 100 * num_passed / float(lang_stats.completed_tests)
                 setattr(lang_stats, f"pass_rate_{i + 1}", pass_rate)
 
+            if lang_stats.test_cases_pass_ratio_count > 0:
+                lang_stats.test_cases_pass_ratio_avg = (
+                    100
+                    * lang_stats.test_cases_pass_ratio_sum
+                    / lang_stats.test_cases_pass_ratio_count
+                )
+            else:
+                lang_stats.test_cases_pass_ratio_avg = 0.0
+
+            del lang_stats.test_cases_pass_ratio_sum
+            del lang_stats.test_cases_pass_ratio_count
+
             # Then format attributes into ready-to-print strings
             for attr in lang_stats.__dict__:
                 val = getattr(lang_stats, attr)
@@ -870,6 +897,11 @@ def summarize_results(results_dir, verbose, stats_languages=None):
     console.rule()
 
     # print(json.dumps(vars(res), indent=4, sort_keys=True))
+    if hasattr(res, "test_cases_pass_ratio_sum"):
+        del res.test_cases_pass_ratio_sum
+    if hasattr(res, "test_cases_pass_ratio_count"):
+        del res.test_cases_pass_ratio_count
+
     return res
 
 
@@ -958,7 +990,7 @@ async def run_test_real(
 
     from cecli import models
     from cecli.coders import Coder
-    from cecli.helpers.conversation import ConversationFiles, ConversationManager
+    from cecli.helpers.conversation import ConversationService
     from cecli.io import InputOutput
     from cecli.main import SwitchCoderSignal
 
@@ -1156,6 +1188,7 @@ async def run_test_real(
             verbose=verbose,
             yes_always_commands=True,
             max_reflections=0,
+            file_diffs=False,
         ),
         map_mul_no_files=4,
         mcp_manager=None,
@@ -1178,13 +1211,12 @@ async def run_test_real(
     dur = 0
     test_outcomes = []
 
-    ConversationManager.initialize(
-        coder,
+    ConversationService.get_manager(coder).initialize(
         reset=True,
         reformat=True,
     )
 
-    ConversationFiles.reset()
+    ConversationService.get_files(coder).reset()
 
     for i in range(tries):
         start = time.time()

cecli_dev-0.98.0/benchmark/variations.1.sh

@@ -0,0 +1,203 @@
+#!/bin/bash
+# Benchmark runner script for testing multiple OpenRouter models
+# Usage: ./run_benchmark_variations.sh [OPTIONS]
+
+set -e  # Exit on error
+
+# Default values
+BASE_NAME="cecli-little-guys-h6"
+EDIT_FORMAT="hashline"
+MAP_TOKENS="512"
+THREADS="1"
+LANGUAGES="javascript,python,rust,go,java"
+HASH_RE="^.[15ef]"
+NUM_TESTS="72"
+EXERCISES_DIR="polyglot-benchmark"
+OUTPUT_DIR="tmp.benchmarks"
+SLEEP_BETWEEN=30  # Seconds to sleep between runs
+
+# List of models to test
+# RERUN
+#    "openrouter/minimax/minimax-m2.1"
+#    "openrouter/qwen/qwen3-vl-235b-a22b-thinking"
+MODELS=(
+#    "openrouter/qwen/qwen3.5-35b-a3b"
+#    "openrouter/xiaomi/mimo-v2-flash"
+#    "openrouter/moonshotai/kimi-k2.5"
+#    "openrouter/minimax/minimax-m2.5"       
+    "openrouter/anthropic/claude-haiku-4.5"
+    "openrouter/openai/gpt-oss-120b"
+    "openrouter/openai/gpt-5-mini"
+    "openrouter/google/gemini-3-flash-preview"
+    "openrouter/deepseek/deepseek-v3.2-exp"
+)
+
+# Parse command line arguments
+while [[ $# -gt 0 ]]; do
+    case $1 in
+        --base-name)
+            BASE_NAME="$2"
+            shift 2
+            ;;
+        --edit-format)
+            EDIT_FORMAT="$2"
+            shift 2
+            ;;
+        --map-tokens)
+            MAP_TOKENS="$2"
+            shift 2
+            ;;
+        --threads)
+            THREADS="$2"
+            shift 2
+            ;;
+        --hash-re)
+            HASH_RE="$2"
+            shift 2
+            ;;
+        --num-tests)
+            NUM_TESTS="$2"
+            shift 2
+            ;;
+        --exercises-dir)
+            EXERCISES_DIR="$2"
+            shift 2
+            ;;
+        --output-dir)
+            OUTPUT_DIR="$2"
+            shift 2
+            ;;
+        --sleep)
+            SLEEP_BETWEEN="$2"
+            shift 2
+            ;;
+        --help)
+            echo "Usage: $0 [OPTIONS]"
+            echo ""
+            echo "Options:"
+            echo "  --base-name NAME      Base name for benchmark runs (default: $BASE_NAME)"
+            echo "  --edit-format FORMAT  Edit format to use (default: $EDIT_FORMAT)"
+            echo "  --map-tokens TOKENS   Map tokens (default: $MAP_TOKENS)"
+            echo "  --threads N           Number of threads (default: $THREADS)"
+            echo "  --hash-re REGEX       Hash regex filter (default: $HASH_RE)"
+            echo "  --num-tests N         Number of tests to run (default: $NUM_TESTS)"
+            echo "  --exercises-dir DIR   Exercises directory (default: $EXERCISES_DIR)"
+            echo "  --output-dir DIR      Output directory (default: $OUTPUT_DIR)"
+            echo "  --sleep SECONDS       Sleep between runs in seconds (default: $SLEEP_BETWEEN)"
+            echo "  --help                Show this help message"
+            echo ""
+            echo "Example:"
+            echo "  $0 --threads 2 --num-tests 5"
+            exit 0
+            ;;
+        *)
+            echo "Unknown option: $1"
+            echo "Use --help for usage information"
+            exit 1
+            ;;
+    esac
+done
+
+# Function to run a single benchmark
+run_benchmark() {
+    local model="$1"
+    local run_name="$2"
+
+    echo "========================================================================"
+    echo "Starting benchmark: $run_name"
+    echo "Model: $model"
+    echo "Time: $(date)"
+    echo "========================================================================"
+
+    # Create the benchmark command
+    ./benchmark/benchmark.py "$run_name" \
+        --new \
+        --model "$model" \
+        --edit-format "$EDIT_FORMAT" \
+        --map-tokens "$MAP_TOKENS" \
+        --threads "$THREADS" \
+        --hash-re "$HASH_RE" \
+        --num-tests "$NUM_TESTS" \
+        --languages "$LANGUAGES" \
+        --tries 2 \
+        --exercises-dir "$EXERCISES_DIR"
+
+    echo "Benchmark completed: $run_name"
+    echo "Results directory: $OUTPUT_DIR/$(ls -t $OUTPUT_DIR | grep "$run_name" | head -1)"
+    echo ""
+}
+
+# Function to generate statistics for all completed runs
+generate_stats() {
+    echo "========================================================================"
+    echo "Generating statistics for all completed runs"
+    echo "========================================================================"
+
+    for dir in "$OUTPUT_DIR"/*; do
+        if [ -d "$dir" ] && [ -f "$dir/.cecli.results.json" ]; then
+            echo "Processing: $(basename "$dir")"
+            ./benchmark/benchmark.py --stats "$dir" || true
+            echo ""
+        fi
+    done
+}
+
+# Main execution
+main() {
+    echo "========================================================================"
+    echo "OpenRouter Model Benchmark Runner"
+    echo "========================================================================"
+    echo "Configuration:"
+    echo "  Base name:      $BASE_NAME"
+    echo "  Edit format:    $EDIT_FORMAT"
+    echo "  Map tokens:     $MAP_TOKENS"
+    echo "  Threads:        $THREADS"
+    echo "  Hash regex:     $HASH_RE"
+    echo "  Num tests:      $NUM_TESTS"
+    echo "  Exercises dir:  $EXERCISES_DIR"
+    echo "  Output dir:     $OUTPUT_DIR"
+    echo "  Sleep between:  ${SLEEP_BETWEEN}s"
+    echo "  Models to test: ${#MODELS[@]}"
+    echo ""
+
+    # Create output directory if it doesn't exist
+    mkdir -p "$OUTPUT_DIR"
+
+    # Run benchmarks for each model
+    for model in "${MODELS[@]}"; do
+        # Create a run name by replacing slashes with hyphens
+        local model_slug=$(echo "$model" | sed 's|/|-|g')
+        local run_name="${BASE_NAME}-${model_slug}"
+
+        run_benchmark "$model" "$run_name"
+
+        # Sleep between runs to avoid rate limiting
+        if [ "$SLEEP_BETWEEN" -gt 0 ]; then
+            echo "Sleeping for ${SLEEP_BETWEEN} seconds before next run..."
+            sleep "$SLEEP_BETWEEN"
+            echo ""
+        fi
+    done
+
+    # Generate statistics
+    generate_stats
+
+    echo "========================================================================"
+    echo "All benchmarks completed!"
+    echo "========================================================================"
+    echo ""
+    echo "Summary of results directories:"
+    ls -la "$OUTPUT_DIR" | grep "$BASE_NAME"
+    echo ""
+    echo "To view statistics for a specific run:"
+    echo "  ./benchmark/benchmark.py --stats $OUTPUT_DIR/<run-directory>"
+    echo ""
+    echo "To compare all results:"
+    echo "  for dir in $OUTPUT_DIR/*$BASE_NAME*; do"
+    echo "    echo \"=== \$(basename \$dir) ===\""
+    echo "    ./benchmark/benchmark.py --stats \"\$dir\" 2>/dev/null | grep -E '(pass_rate|total_cost|completed_tests)' || true"
+    echo "  done"
+}
+
+# Run main function
+main

cecli_dev-0.98.0/benchmark/variations.2.sh

@@ -0,0 +1,203 @@
+#!/bin/bash
+# Benchmark runner script for testing multiple OpenRouter models
+# Usage: ./run_benchmark_variations.sh [OPTIONS]
+
+set -e  # Exit on error
+
+# Default values
+BASE_NAME="cecli-little-guys-h6"
+EDIT_FORMAT="hashline"
+MAP_TOKENS="512"
+THREADS="1"
+LANGUAGES="javascript,python,rust,go,java"
+HASH_RE="^.[15ef]"
+NUM_TESTS="72"
+EXERCISES_DIR="polyglot-benchmark"
+OUTPUT_DIR="tmp.benchmarks"
+SLEEP_BETWEEN=30  # Seconds to sleep between runs
+
+# List of models to test
+# RERUN
+#    "openrouter/minimax/minimax-m2.1"
+#    "openrouter/qwen/qwen3-vl-235b-a22b-thinking"
+MODELS=(
+    "openrouter/qwen/qwen3.5-35b-a3b"
+    "openrouter/xiaomi/mimo-v2-flash"
+    "openrouter/moonshotai/kimi-k2.5"
+    "openrouter/minimax/minimax-m2.5"       
+#    "openrouter/anthropic/claude-haiku-4.5"
+#    "openrouter/openai/gpt-oss-120b"
+#    "openrouter/openai/gpt-5-mini"
+#    "openrouter/google/gemini-3-flash-preview"
+#    "openrouter/deepseek/deepseek-v3.2-exp"
+)
+
+# Parse command line arguments
+while [[ $# -gt 0 ]]; do
+    case $1 in
+        --base-name)
+            BASE_NAME="$2"
+            shift 2
+            ;;
+        --edit-format)
+            EDIT_FORMAT="$2"
+            shift 2
+            ;;
+        --map-tokens)
+            MAP_TOKENS="$2"
+            shift 2
+            ;;
+        --threads)
+            THREADS="$2"
+            shift 2
+            ;;
+        --hash-re)
+            HASH_RE="$2"
+            shift 2
+            ;;
+        --num-tests)
+            NUM_TESTS="$2"
+            shift 2
+            ;;
+        --exercises-dir)
+            EXERCISES_DIR="$2"
+            shift 2
+            ;;
+        --output-dir)
+            OUTPUT_DIR="$2"
+            shift 2
+            ;;
+        --sleep)
+            SLEEP_BETWEEN="$2"
+            shift 2
+            ;;
+        --help)
+            echo "Usage: $0 [OPTIONS]"
+            echo ""
+            echo "Options:"
+            echo "  --base-name NAME      Base name for benchmark runs (default: $BASE_NAME)"
+            echo "  --edit-format FORMAT  Edit format to use (default: $EDIT_FORMAT)"
+            echo "  --map-tokens TOKENS   Map tokens (default: $MAP_TOKENS)"
+            echo "  --threads N           Number of threads (default: $THREADS)"
+            echo "  --hash-re REGEX       Hash regex filter (default: $HASH_RE)"
+            echo "  --num-tests N         Number of tests to run (default: $NUM_TESTS)"
+            echo "  --exercises-dir DIR   Exercises directory (default: $EXERCISES_DIR)"
+            echo "  --output-dir DIR      Output directory (default: $OUTPUT_DIR)"
+            echo "  --sleep SECONDS       Sleep between runs in seconds (default: $SLEEP_BETWEEN)"
+            echo "  --help                Show this help message"
+            echo ""
+            echo "Example:"
+            echo "  $0 --threads 2 --num-tests 5"
+            exit 0
+            ;;
+        *)
+            echo "Unknown option: $1"
+            echo "Use --help for usage information"
+            exit 1
+            ;;
+    esac
+done
+
+# Function to run a single benchmark
+run_benchmark() {
+    local model="$1"
+    local run_name="$2"
+
+    echo "========================================================================"
+    echo "Starting benchmark: $run_name"
+    echo "Model: $model"
+    echo "Time: $(date)"
+    echo "========================================================================"
+
+    # Create the benchmark command
+    ./benchmark/benchmark.py "$run_name" \
+        --new \
+        --model "$model" \
+        --edit-format "$EDIT_FORMAT" \
+        --map-tokens "$MAP_TOKENS" \
+        --threads "$THREADS" \
+        --hash-re "$HASH_RE" \
+        --num-tests "$NUM_TESTS" \
+        --languages "$LANGUAGES" \
+        --tries 2 \
+        --exercises-dir "$EXERCISES_DIR"
+
+    echo "Benchmark completed: $run_name"
+    echo "Results directory: $OUTPUT_DIR/$(ls -t $OUTPUT_DIR | grep "$run_name" | head -1)"
+    echo ""
+}
+
+# Function to generate statistics for all completed runs
+generate_stats() {
+    echo "========================================================================"
+    echo "Generating statistics for all completed runs"
+    echo "========================================================================"
+
+    for dir in "$OUTPUT_DIR"/*; do
+        if [ -d "$dir" ] && [ -f "$dir/.cecli.results.json" ]; then
+            echo "Processing: $(basename "$dir")"
+            ./benchmark/benchmark.py --stats "$dir" || true
+            echo ""
+        fi
+    done
+}
+
+# Main execution
+main() {
+    echo "========================================================================"
+    echo "OpenRouter Model Benchmark Runner"
+    echo "========================================================================"
+    echo "Configuration:"
+    echo "  Base name:      $BASE_NAME"
+    echo "  Edit format:    $EDIT_FORMAT"
+    echo "  Map tokens:     $MAP_TOKENS"
+    echo "  Threads:        $THREADS"
+    echo "  Hash regex:     $HASH_RE"
+    echo "  Num tests:      $NUM_TESTS"
+    echo "  Exercises dir:  $EXERCISES_DIR"
+    echo "  Output dir:     $OUTPUT_DIR"
+    echo "  Sleep between:  ${SLEEP_BETWEEN}s"
+    echo "  Models to test: ${#MODELS[@]}"
+    echo ""
+
+    # Create output directory if it doesn't exist
+    mkdir -p "$OUTPUT_DIR"
+
+    # Run benchmarks for each model
+    for model in "${MODELS[@]}"; do
+        # Create a run name by replacing slashes with hyphens
+        local model_slug=$(echo "$model" | sed 's|/|-|g')
+        local run_name="${BASE_NAME}-${model_slug}"
+
+        run_benchmark "$model" "$run_name"
+
+        # Sleep between runs to avoid rate limiting
+        if [ "$SLEEP_BETWEEN" -gt 0 ]; then
+            echo "Sleeping for ${SLEEP_BETWEEN} seconds before next run..."
+            sleep "$SLEEP_BETWEEN"
+            echo ""
+        fi
+    done
+
+    # Generate statistics
+    generate_stats
+
+    echo "========================================================================"
+    echo "All benchmarks completed!"
+    echo "========================================================================"
+    echo ""
+    echo "Summary of results directories:"
+    ls -la "$OUTPUT_DIR" | grep "$BASE_NAME"
+    echo ""
+    echo "To view statistics for a specific run:"
+    echo "  ./benchmark/benchmark.py --stats $OUTPUT_DIR/<run-directory>"
+    echo ""
+    echo "To compare all results:"
+    echo "  for dir in $OUTPUT_DIR/*$BASE_NAME*; do"
+    echo "    echo \"=== \$(basename \$dir) ===\""
+    echo "    ./benchmark/benchmark.py --stats \"\$dir\" 2>/dev/null | grep -E '(pass_rate|total_cost|completed_tests)' || true"
+    echo "  done"
+}
+
+# Run main function
+main

cecli_dev-0.97.4/benchmark/primary_variations.sh → cecli_dev-0.98.0/benchmark/variations.3.sh

@@ -5,7 +5,7 @@
 set -e  # Exit on error
 
 # Default values
-BASE_NAME="cecli-base-d-big-3"
+BASE_NAME="cecli-little-guys-d6"
 EDIT_FORMAT="diff"
 MAP_TOKENS="512"
 THREADS="1"
@@ -21,14 +21,15 @@ SLEEP_BETWEEN=30  # Seconds to sleep between runs
 #    "openrouter/minimax/minimax-m2.1"
 #    "openrouter/qwen/qwen3-vl-235b-a22b-thinking"
 MODELS=(
-#    "openrouter/deepseek/deepseek-v3.2-exp"
-    "openrouter/moonshotai/kimi-k2.5"
+#    "openrouter/qwen/qwen3.5-35b-a3b"
+#    "openrouter/xiaomi/mimo-v2-flash"
+#    "openrouter/moonshotai/kimi-k2.5"
+#    "openrouter/minimax/minimax-m2.5"       
+    "openrouter/anthropic/claude-haiku-4.5"
     "openrouter/openai/gpt-oss-120b"
-    "openrouter/openai/gpt-5.2"   
+    "openrouter/openai/gpt-5-mini"
     "openrouter/google/gemini-3-flash-preview"
-    "openrouter/google/gemini-3-pro-preview"
-    "openrouter/anthropic/claude-haiku-4.5"
-    "openrouter/anthropic/claude-sonnet-4.5" 
+    "openrouter/deepseek/deepseek-v3.2-exp"
 )
 
 # Parse command line arguments
@@ -118,6 +119,7 @@ run_benchmark() {
         --hash-re "$HASH_RE" \
         --num-tests "$NUM_TESTS" \
         --languages "$LANGUAGES" \
+        --tries 2 \
         --exercises-dir "$EXERCISES_DIR"
 
     echo "Benchmark completed: $run_name"