explainx.ai350K+ learn from us
TrendingAI News TodayExplainx for Business
Pricing
explainx.ai

Upskill in AI — 16 free pathways, live workshops & bootcamps, and 50+ courses from practitioners. Plus the skills, tools, and MCP servers to practice on.

follow us

follow on google

Add explainx.ai as a preferred source

corporate training

support@explainx.ai

get started

Find your pathTake Free Evaluation

community

Join the community

learn

mind: share how you thinkpathways — start freeworkshopsbootcampscoursescompare Explainxcertificationsmock testsexplainx universitycorporate traininglearn skills & mcp

discover

skillsmcp serversexplainx mcpestacktoolsmdx readeragentsllmsdesignsdictionarypeopleagi trackerfelony benchranks

company

aboutvisionmissionteaminstructorsteach on explainxpartnershipscommunityhackathonscareers

content

daily AI newsstate of AI — live resultsblogreleasespromptsgeneratorsresource libraryfor LLMsexplainx.ai kids

solutions

all solutionsdeveloper upskillingmarketing upskillingproduct manager upskillingleadership upskilling

newsletter · weekly

Get AI news, tools, and insights in your inbox.

supportcontactprivacytermsdata rightshow we create contentsubmission guidelines

© 2026 AISOLO Technologies Pvt Ltd

explainx.ai

On this page

  • TL;DR: verified numbers from the model card
  • Why a separate OpenCode adapter exists
  • What changed on SWE-bench Pro hard 51
  • Thinking off is the product setting
  • How to reproduce the harness (vLLM + OpenCode 2.2.7)
  • Evaluation caveats the card discloses
  • Economics: turns and tokens beat vanity percentages
  • How this fits export-control and cloud planning
  • OpenCode versus Pi: when to load which LoRA
  • Pairing with GGUF and Apple Silicon paths
  • safeagents and reward hacking during RL
  • Procurement language for vendor bake-offs
  • Failure modes we expect in production
  • Cost worksheet (illustrative)
  • Incident response when benchmarks drive deadlines
  • What is coming next from harness-hub
  • Related reading
  • Bottom line
← Back to blog

explainx / blog

Qwen3.8-27B OpenCode Adapter: 51% SWE-Bench Pro at 6–8× Fewer Tokens

Qwen, OpenCode, Coding Agents, Hugging Face, Open Models

Part of Open-Weight Models

harness-hub released Qwen3.8-27B-opencode-adapted: SWE-bench Pro hard-51 rises from 35.3% to 51.0% under OpenCode 2.2.7 with far fewer turns. vLLM and config steps inside.

Nov 18, 2026·8 min read·Yash Thakker
add explainx.ai
go deep
Qwen3.8-27B OpenCode Adapter: 51% SWE-Bench Pro at 6–8× Fewer Tokens

harness-hub published a LoRA adapter that makes Qwen3.8-27B behave like an OpenCode-native coder—not a generic chat model with tools bolted on. The Qwen3.8-27B-opencode-adapted card, updated November 18, 2026, reports 51.0% pass@1 on SWE-bench Pro (hard 51) with thinking off, up from 35.3% for the base checkpoint in the same OpenCode harness. Median generated tokens per solved task fall to roughly one-sixth of what the base model spends when thinking is enabled.

explainx.ai covered the Pi sibling in Qwen3.8-27B SWE-bench Pro 63% on Pi and the broader harness-engineering tier debate. This post is the OpenCode-specific install guide and score reconciliation so you do not compare OpenCode hard-51 rows with Alibaba Claude Code full-bench marketing.

Daily briefFrom Yash · 350K+ students taught

Catch up on AI

One AI story explained in plain English, plus one thing to try. Every day, free. Unsubscribe anytime.

TL;DR: verified numbers from the model card

table · 3 cols
BenchmarkBase Qwen3.8-27B (OpenCode)+ OpenCode LoRA
SWE-bench Pro hard 5135.3%51.0%
Terminal-Bench 2.1 (84 tasks)46.4%52.4%
Median turns on shared SWE solves4813
Median tokens (SWE solves vs base + thinking)33.5k~3.9k

All rows assume one attempt per task, 100 agent steps, text-only serving (vision tower off), temperature 0.8, top_p 1.0, and thinking off unless the card explicitly compares thinking runs.

Why a separate OpenCode adapter exists

OpenCode, Pi, Hermes, and Claude Code expose different tool schemas, compaction rules, and permission flows. The harness-hub org argues you should keep the harness and adapt the model—train a small LoRA per harness and hot-swap at runtime. That is the same design pattern Liquid LFM adapters pursue on smaller bases, but here the target is a 27B open-weight checkpoint teams already rent or self-host.

Training stack pieces the org README links:

  • AsyncGRPO in TRL with LoRA
  • Repo2RLEnv tasks built from real agent trajectories
  • safeagents proxy blocking reward-hacking tool calls during rollouts

If you run multi-harness shops—OpenCode for one team, Pi for another—one vLLM server can host multiple LoRA modules as long as each CLI points at the correct alias.

What changed on SWE-bench Pro hard 51

OpenCode harness benchmark results chart for Qwen3.8-27B LoRA adapter on SWE-bench ProOpenCode harness benchmark results chart for Qwen3.8-27B LoRA adapter on SWE-bench Pro Image: harness-hub

The card breaks outcomes into newly solved, lost, and sign-test p-values:

table · 2 cols
MetricValue
Newly solved tasks11
Lost tasks3
Difference+15.7 ± 13.9 points
Sign testp = 0.057 (borderline on 51 tasks)

On Terminal-Bench 2.1, the adapter gains +6.0 points on all 84 graded tasks, but the card highlights +13.6 points on the 66 tasks not used to pick the checkpoint (p = 0.035) as the stronger generalization signal.

Checkpoint selection caveat. Step 150 was chosen using SWE-bench Pro plus an 18-task Terminal-Bench subset watched during training—so headline benchmarks were not fully blind. Treat gains as real for OpenCode users, but run your own repo slice before you reorganize spend.

Thinking off is the product setting

The adapter was trained with thinking off. The card shows that enabling xhigh thinking on the adapter does not materially beat thinking-off accuracy while multiplying tokens.

table · 3 cols
Mode on hard 51Adapter scoreMedian generated tokens (illustrative row)
Base + thinking47.1%31.8k
Adapter, thinking off51.0%~3.9k

Practical takeaway: serve with enable_thinking: false in chat template kwargs, exactly as the vLLM snippet documents.

How to reproduce the harness (vLLM + OpenCode 2.2.7)

1. Serve LoRA with vLLM

From the model card:

bash
vllm serve Qwen/Qwen3.8-27B \
  --enable-lora --max-lora-rank 32 \
  --lora-modules qwen3.8-opencode=harness-hub/Qwen3.8-27B-opencode-adapted \
  --language-model-only --enable-prefix-caching --mamba-cache-mode align \
  --enable-auto-tool-choice --tool-call-parser qwen3_xml --reasoning-parser qwen3 \
  --default-chat-template-kwargs '{"enable_thinking": false}' \
  --max-model-len 131072

Prefix caching on Qwen3.8 hybrid stacks needs --mamba-cache-mode align. Without it, every turn re-reads full history and agent bills explode.

2. Install OpenCode and wire the provider

bash
npm install -g @opencode/cli@2.2.7

Add to ~/.config/opencode/opencode.json:

json
{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "local": {
      "npm": "@ai-sdk/openai-compatible",
      "options": {
        "baseURL": "http://localhost:8000/v1",
        "apiKey": "dummy"
      },
      "models": {
        "qwen3.8-opencode": {}
      }
    }
  }
}

Run a smoke task:

bash
opencode run --standalone --model local/qwen3.8-opencode "Find out why the tests in tests/test_parser.py fail and fix the bug."

Cross-check OpenCode versus other terminals if your team mixes CLIs.

Evaluation caveats the card discloses

Builders should read the limitations section before tweeting leaderboard screenshots:

table · 2 cols
CaveatWhy it matters
Text-only serverOpenCode may request web search or treat files as images; disabled vision and search ended some Terminal-Bench rollouts early
Tunnel outagesPublic eval infra dropped sessions; maintainers reran affected trajectories (base SWE score moved from 31.4% pre-rerun to 35.3%)
Test file editsAgents sometimes wrote scratch tests; official graders reset tests before grading
Not comparable to Qwen cardAlibaba numbers use other harnesses, longer budgets, and thinking on

Those disclaimers are why explainx.ai keeps claimed-vs-verified posts separate per harness.

Economics: turns and tokens beat vanity percentages

Generated tokens per task comparing Qwen3.8-27B base versus OpenCode LoRA adapterGenerated tokens per task comparing Qwen3.8-27B base versus OpenCode LoRA adapter Image: harness-hub

On tasks both base and adapter solve on SWE-bench Pro hard 51, medians are 13 vs 48 turns and 1.9k vs 6.8k generated tokens. For a 10k-issue backlog, that is the difference between buying another GPU and shipping with the same cluster.

Pair cost math with right-sized open inference APIs and vLLM routing posts when you negotiate provider contracts.

How this fits export-control and cloud planning

OpenCode adapters do not change EAR rules, but they change how much foreign-hosted GPU you need. Teams shrinking tokens per merge may keep training on domestic Ascend pools while US-hosted frontier APIs serve only the hardest tasks—relevant as BIS remote-access licensing reshapes Southeast Asia rentals.

OpenCode versus Pi: when to load which LoRA

harness-hub now ships two Qwen3.8-27B adapters:

table · 4 cols
AdapterHard-51 SWE (card)Median turns (shared solves)CLI
Pi62.7%lower on Pi traces@earendil-works/pi-coding-agent
OpenCode51.0%13 vs 48 on SWE@opencode/cli@2.2.7

Do not mount the OpenCode LoRA behind Pi or vice versa—the tool XML schemas differ. If your org standardized on OpenCode because of multi-model routing experiments, this release is the first official Qwen-native path with published eval logs.

Teams running both CLIs on one vLLM pod can register:

bash
--lora-modules qwen3.8-pi=harness-hub/Qwen3.8-27B-pi-adapted qwen3.8-opencode=harness-hub/Qwen3.8-27B-opencode-adapted

…then point each harness at the correct model id.

Pairing with GGUF and Apple Silicon paths

The org also publishes Qwen3.8-27B-opencode-adapted-GGUF companions for llama.cpp, mirroring the Pi release. The card warns GGUF bases drift from BF16 training weights—Q8_0 is the suggested default, but re-evaluate before you quote 51% on Metal.

For Mac-heavy teams compare against llama.cpp release cadence and TensorFold local inference if you need sub-27B fallbacks when VRAM is tight.

safeagents and reward hacking during RL

The harness-hub blog positions safeagents as a 10–20 ms rules engine that zeros rollouts when models attempt shortcut tool abuse during AsyncGRPO. That matters because coding-agent felony bench tracking shows real-world agents already publish malicious packages when eval guardrails are weak.

If you fork the recipe:

  1. Keep safeagents (or equivalent) in the training loop, not only production.
  2. Log blocked tool calls per checkpoint—spikes mean the LoRA is gaming rewards.
  3. Mirror MCP misbinding defenses in prod after you ship.

Procurement language for vendor bake-offs

Paste this into RFPs when a vendor quotes SWE-bench Pro:

Scores must name harness name, @opencode/cli version, thinking on/off, step budget, and LoRA commit SHA. OpenCode numbers are not interchangeable with Claude Code rows on vendor model cards.

That single paragraph prevents 61.7% marketing slides from sneaking into OpenCode pilots without re-run costs.

Failure modes we expect in production

table · 3 cols
SymptomLikely causeFix
Tool calls ignoredWrong chat template or thinking left onForce enable_thinking: false
Sessions die on imagesVision requests against text-only serverEnable vision tower or strip image tools
Runaway spendMissing prefix cache align on hybrid QwenAdd --mamba-cache-mode align
Scores collapse after upgradeOpenCode >2.2.7 behavior changePin CLI + rerun 20-task slice

Cost worksheet (illustrative)

Assume $2/M input and $6/M output list pricing on a hosted 27B endpoint (your contract may differ):

  1. Median 1.9k output tokens per solved SWE task on the adapter (card, thinking off).
  2. Multiply by issues per week your agent fleet accepts.
  3. Compare against baseline 6.8k tokens on shared solves—adapter wins when output dominates the bill.

Add GPU amortization if self-hosting: a single 48 GB card serving LoRA often beats frontier API pricing when utilization stays above 60%, but only if legal can approve export-category hardware in your region.

Incident response when benchmarks drive deadlines

When leadership sees +15 points on Twitter and demands next-day rollout:

  1. Publish an internal dashboard with harness version, LoRA SHA, and cost per merged PR.
  2. Run 20 private issues before you touch customer repos.
  3. Keep rollback weights (base Qwen3.8-27B without LoRA) one config flag away.

That playbook mirrors Semianalysis agent failure rates—speed without harness discipline creates incidents, not velocity.

What is coming next from harness-hub

The Qwen3.8-27B collection page still lists Pi (OpenCode and Hermes coming)—OpenCode now shipped; Hermes may follow the same AsyncGRPO recipe. The org promises a full training blog on Hugging Face; until it lands, fork TRL async_grpo_harbor if you need a custom harness adapter for an internal CLI.

Related reading

  • Qwen3.8-27B Pi adapter claimed vs verified
  • Agent harness tier list reactions
  • Higher SWE scores, lower merge rates
  • External sandboxes for coding agents
  • Semianalysis agents fail benchmarks

Bottom line

OpenCode users now have a verified LoRA with public vLLM commands—not a rumor screenshot. Pin CLI 2.2.7, keep thinking off, and rerun your issue slice before you rebalance API spend away from frontier models.

Benchmarks and commands follow the harness-hub model card accessed November 18, 2026. Rerun evals after you upgrade OpenCode beyond 2.2.7.

Spotted something out of date? Let us know.
Daily briefFrom Yash · 350K+ students taught

Catch up on AI

One AI story explained in plain English, plus one thing to try. Every day, free. Unsubscribe anytime.

  • ✓ Free
  • ✓ One email a day, no spam
  • ✓ Unsubscribe anytime
Yash Thakker

Written by

Yash Thakker

Yash is an AI expert with over 300K learners. Join his workshops →

View Yash Thakker in People in AI →

Related posts

Nov 17, 2026

Qwen3.8-27B SWE-Bench Pro 63% on Pi: Claimed vs Official 61.7%

A November 2026 wave of posts claimed Qwen3.8-27B hits 63% on SWE-bench Pro using a new Pi harness adapter. The official Hugging Face model card still reports 61.7% evaluated with the Claude Code harness at temp 1.0—harness choice moves the number.

Nov 11, 2026

Qwen4-Max Preview Lands on Alibaba Token Plan: What Builders Can Try Today

On November 11, 2026 Alibaba’s Qwen team said Qwen4 is arriving and put Qwen4-Max into preview on Token Plan and four consumer apps. There is still no public weight drop, API list price row, or full four-model rollout—but builders with a Token Plan can call the preview id today.

Nov 1, 2026

FineDNA and Carbon-X: Hugging Face Open-Sources a DNA Dataset and a 14B Model

Hugging Face Bio published FineDNA, the full training data behind Carbon-X, a Qwen3-14B model taught to read DNA. This post covers what is in the dataset, the benchmark table versus frontier models, the cost claims, and the limits.