Agentic Coding
Archived

Agentic Coding

A benchmark for evaluating AI models on complex, real-world coding tasks that require multi-step reasoning, tool use, and autonomous problem-solving.
Overview

The Snorkel Agentic Coding benchmark comprises 100 multi-step coding tasks, evenly distributed across four difficulty tiers, designed to evaluate models across a diverse range of capabilities germane to real-world software engineering work.

Taking insights from our contributions to the Terminal-Bench project, our Agentic Coding tasks evaluate agents in fully sandboxed execution environments. Each task is paired with a human-validated reference solution, comprehensive unit tests, and scoring rubrics that assess both final outputs and the
agent's trajectory.

Leaderboard

Rank Model Score
1 Claude Opus 4.6
65.2%
2 Claude Opus 4.5
58%
3 Claude Sonnet 4.5
57.6%
4 Gemini 3 Pro Preview
51.6%
5 gpt-5.2
49.4%
6 gpt-5
45.2%
7 Kimi-K2-Thinking
36.8%
8 Devstral 2
33.2%
9 Grok 4.1 Fast
25.2%
10 Qwen 3 Coder 480B
18.8%
11 Mistral Large 3
13.8%

Sample Task

Incident Commander: Payments Canary Rollback

You are the on-call SRE responding to a spike in 5xx errors after a canary rollout of payments-api. All observability artifacts, runbooks, and config files are already packaged inside the container. No network access is allowed.

Requirements

  1. Diagnose the incident using logs, metrics, and traces to identify the root cause and blast radius.

  2. Execute mitigation per the runbook (/app/runbooks/payments-canary-rollback.md):

    • Update /app/config/service_state.json as specified in the runbook

    • Only modify payments-api - do not change any other services

    • Only change fields specified in the runbook - preserve all other fields exactly

    • Create any required configuration files specified in the runbook

  3. Produce output files documenting your diagnosis, actions, and validation.

Inputs

  • /app/data/logs/app.log — application logs

  • /app/data/metrics/before.json — pre-incident metrics

  • /app/data/metrics/after.json — post-mitigation metrics

  • /app/data/traces/trace_sample.json — distributed traces

  • /app/data/deployments/ — deployment manifests

  • /app/runbooks/payments-canary-rollback.md — mitigation runbook

  • /app/config/service_state.json — service configuration

Outputs

All outputs must be created exactly as specified below.

/app/outputs/incident_summary.md

Markdown document with these sections: Overview, Root Cause, Blast Radius, Evidence, Actions Taken, Validation, Next Steps.

  • Cite full artifact paths used in analysis

  • Include pod, DB host, and failure mode in root cause

  • Include the primary trace ID (the first payments-api canary error trace with DB connection issues, by timestamp), rollback_token, and correlation ID

  • List affected transaction IDs and count

  • Include the incident time window (start and end timestamps of payments-api canary DB errors)

  • Confirm unaffected services explicitly - check all other services in service_state.json (excluding payments-api) and confirm they are unaffected

  • Show before/after metrics with deltas

  • Include an "SLO Budget" section containing the words "SLO" and "budget", plus the calculated remaining budget value (monthly_budget_pct minus consumed_pct, rounded to 2 decimal places)

  • Use "disable" or "disabled" when describing the canary action

/app/outputs/actions.json

JSON array with exactly FIVE action objects:

  1. {"action": "disable_canary", "target": "payments-api", "status": "...", "details": "...", "evidence": "...", "rollback_token": "...", "correlation_id": "..."}

    • evidence must reference app.log and include the primary trace ID

    • correlation_id must be extracted from the relevant log entries

  2. {"action": "create_alert_suppression", "target": "payments-api", "status": "...", "details": "..."}

  3. {"action": "create_followup_ticket", "target": "payments-api", "status": "...", "details": "...", "priority": "...", "assigned_team": "..."}

    • priority and assigned_team must be determined per the runbook
  4. {"action": "notify_stakeholders", "target": "payments-api", "status": "...", "details": "...", "channel": "...", "escalation_level": "..."}

    • channel and escalation_level must be determined per the runbook
  5. {"action": "update_deployment_status", "target": "payments-api", "status": "...", "details": "...", "previous_status": "...", "new_status": "...", "canary_version": "..."}

    • previous_status: the status from the deployment manifest before rollback

    • new_status: "rolled_back"

    • canary_version: the version from the deployment manifest

/app/outputs/postcheck.json

JSON object with fields: error_rate_pct, p99_latency_ms, validation_passed, notes, affected_txn_count, impact_pct, incident_window, slo_budget, mttr_seconds, error_velocity

  • error_rate_pct and p99_latency_ms: values from after-metrics for payments-api

  • validation_passed: boolean based on runbook thresholds

  • notes: must contain "after.json", the literal metric names ("error_rate_pct", "p99_latency_ms"), and their threshold values ("5" and "400")

  • affected_txn_count: count of unique transactions from payments-api canary DB error log entries

  • impact_pct: (affected_txn_count / total_requests_from_before_metrics) * 100, rounded to 2 decimal places

  • incident_window: object with start and end fields containing the first and last payments-api canary DB error timestamps from logs

  • slo_budget: object with these exact fields:

    • monthly_budget_pct: 0.1 (representing 99.9% availability SLO)

    • consumed_pct: the error_rate_pct from before-metrics (during incident)

    • remaining_pct: monthly_budget_pct minus consumed_pct

  • mttr_seconds: Mean Time To Recovery - seconds between first canary DB error and last canary DB error (parse timestamps and compute difference)

  • error_velocity: errors per minute during incident - affected_txn_count divided by (mttr_seconds / 60), rounded to 2 decimal places

/app/config/alert_suppression.json

JSON object with fields:

  • service: "payments-api"

  • suppression_window_minutes: 30

  • reason: description of the incident

  • created_by: "incident-commander"

  • expires_at: ISO 8601 timestamp, computed as incident start time (first canary error) + 30 minutes

/app/data/deployments/canary_manifest.json

Update the deployment manifest:

  • Find the payments-api deployment entry

  • Change its status field from "active" to "rolled_back"

  • Preserve all other fields and entries exactly

Methodology

METRIC
Pass@5, evaluated through the Harbor evaluation harness.
TIMEOUT
Each task has a specific timeout limit, with an absolute maximum of 30 minutes for both agent and verifier.
ENVIRONMENT
Fully sandboxed execution. Each sample is accompanied by all data and dependencies required, a test suite, a rubric for human and LLM evaluators, and a golden solution.
Difficulty Tiers
Four levels, evenly distributed across 100 tasks. Calibrated to provide signal from efficiency-optimized to frontier-level models.

Behind the benchmark

The current version of the benchmark spans a wide range of task categories, from typical software engineering related tasks, to advanced ML and data analytics, as well as build and dependency management tasks, and tests agents on long-horizon planning, tracking tasks, evaluating and executing their own solutions, and recovering from potential errors and incorrect previous steps.

Our benchmark is built to challenge even the most advanced frontier models. Tasks are constructed with experts in the loop, confirming every challenge to be solvable in the environments in which they run, and verifying the reliability of all dependencies. We have calibrated the tasks so they deliver a range of difficulties, providing meaningful feedback for agents and models across the cost/performance spectrum — from those pursuing Pareto-optimal results, to those that are delivering truly frontier-level capabilities.

From the blog

Image for Introducing the Snorkel Agentic Coding Benchmark

Introducing the Snorkel Agentic Coding Benchmark

Today, we’re sharing details about the Snorkel Agentic Coding benchmark—a comprehensive evaluation suite designed to test whether agents can handle...
January 8, 2026

Get notified when we launch a new benchmark

Share this benchmark

More benchmarks

Agentic Coding

Agentic Coding 2.0

Evaluating whether coding agents can plan, execute, verify, and recover across complex terminal-native engineering tasks.

By pass@1
1
Image
GPT-6 Astra
47.6%
2
Image
Fable 5.1
39.6%
3
Image
Opus 5
38.9%
Software Engineering

SWE-bench CLI

Tests whether an AI coding agent can diagnose and deliver a validated, multi-file change in a real open-source repository, navigating code, tests, dependencies, and tooling through the command line.

By pass@1
1
Image
Fable 5.1
14.5%
2
Image
Opus 5
14.0%
3
Image
GPT-6 Astra
12.4%
Enterprise Environments

SnorkelUnderwrite 2.0

Measures an agent’s ability to turn incomplete, distributed insurance evidence into an auditable underwriting decision while respecting authority and policy constraints.

By pass@1
1
Image
GLM 5.3
30.4%
2
Image
DeepSeek V4 Pro
28.5%
3
Image
Kimi K3
27.9%
Enterprise Environments

SnorkelManufacturing

Measures how well AI agents can turn fragmented plant-floor, engineering, and supplier evidence into safe, technically defensible decisions and actions.

By pass@1
1
Image
Grok 4.6
16.4%
2
Image
GLM 5.3
11.9%
3
Image
Fable 5.1
11.3%
Enterprise Environments

SnorkelRevOps

Tests whether AI agents can reconcile revenue systems, enforce hard commercial controls, and carry a decision through to the correct business state.

By pass@1
1
Image
Grok 4.6
15.8%
2
Image
Fable 5.1
14.0%
3
Image
Kimi K3
13.1%
Enterprise Environments

SnorkelFinance 2.0

Scores how agents gather evidence, perform financial analysis, follow compliance constraints, and complete required state updates in a simulated environment.

By pass@1
1
Image
Grok 4.6
25.1%
2
Image
GLM 5.3
21.2%
3
Image
Fable 5.1
19.6%
Enterprise Environments

SnorkelLegal

Evaluates AI agents on whether they can advance a legal matter while preserving the chain from controlling evidence to procedure, authority, and action.

By pass@1
1
Image
Grok 4.6
40.7%
2
Image
GLM 5.3
37.8%
3
Image
Muse Spark 1.3
36.5%
Knowledge Work

WorkplaceAgents

Evaluates AI agents on economically valuable professional work and verifiable workplace deliverables.

By pass@1
1
Image
Muse Spark 1.3
17.7%
2
Image
Grok 4.6
17.1%
3
Image
Fable 5.1
15.8%
Agentic Coding

Terminal-Bench 4.0

A benchmark to measure and evolve with the frontier of agent work: real terminal environments, real software engineering tasks, and a rolling task set that is revised as agents catch up to it.

By Resolution Rate
1
Image
GPT-6 Astra
58.2%
2
Image
Fable 5.1
57.9%
3
Image
Opus 5
53.9%
Scientific & Research Workflows

Terminal-Bench-Science

A benchmark for evaluating AI agents on workflows from researchers’ own work. Scientists, not model developers or vendors, set the bar for scientific capability in AI.

By Resolution Rate
1
Image
Fable 5.1
40.0%
2
Image
Opus 5
30.0%
3
Image
GPT-5.6 Sol
22.4%
Agentic Coding

Terminal-Bench 3.0

The next frontier benchmark for agent work. A harder, more domain-diverse successor to Terminal-Bench 2.1 — built in the open, task by task, under continuous adversarial review.

By Resolution Rate
1
Image
Opus 5
42.7%
2
Image
GPT-5.6 Sol
34.6%
3
Image
Fable 5
34.1%
Computer Use

OSWorld 2.0

Long-horizon professional workflows with verifiable outcomes across 55 sub-industries. 147 public tasks of a 1,500+ task corpus, sourced and validated by 300+ industry experts.

By binary accuracy (500 steps)
1
Image
Opus 5 · max
44.33%
2
Image
Opus 5 · xhigh
36.89%
3
Image
Opus 5 · high
33.33%
Software Engineering

Senior SWE-Bench

Evaluating coding agents on senior-level engineering work.

Tasteful Solve Rate
1
Image
Fable 5.1
34.7%
2
Image
Fable 5
34.7%
3
Image
Opus 5
34.7%
Knowledge Work

Agents’ Last Exam

Evaluating AI agents on long-horizon, economically valuable professional workflows with verifiable outcomes.

By Binary Accuracy
1
Image
GPT-6 Astra
34.2
2
Image
Muse Spark 1.3
32.2%
3
Image
Opus 5
31.6%
Software Engineering

SlopCode Bench

Measures code quality degradation in AI-assisted codebases. Tracks checkpoint solve rates, erosion (code bloat), and verbosity under realistic repo conditions.

Top Models by Iso Solve
1
Image
GPT-5.5
28.06%
2
Image
GPT-5.3-Codex
26.02%
3
Image
GPT-5.4
23.47%
Capability/Efficiency

Continual Learning Bench

Evaluates whether AI systems improve from prior experience across sequential, stateful tasks, measuring real in-context learning, not just raw capability.

Top Systems (Agg. Reward)
1
Image
Sonnet 4.6 · ICL
+0.196
2
Image
GPT-5.4 · ICL
+0.189
3
Image
Sonnet 4.6 · Claude Code
+0.185
of

For models that need to be right. Not just good enough.