June 30, 2026 — 4:19 PM: ModelScope announced Agents-A1 on X — a 35B MoE agentic model from InternScience built for long-horizon search, engineering, scientific research, instruction following, and tool calling. Weights landed on Hugging Face the same day under Apache 2.0, with a technical report claiming trillion-parameter-class agent performance without trillion-parameter weights.
The launch sits in a crowded week: LongCat-2.0 from Meituan, ongoing Qwen 3.6 local-dev hype, and Fable 5 still offline. Agents-A1's pitch is different — not raw coding SWE scores alone, but heterogeneous agent horizons: search loops, science tools, instruction evals, and function-calling at 256K context.
TL;DR — what people asked on X
| Question | Answer |
|---|---|
| What is it? | 35.11B MoE agent model, qwen3_5_moe architecture, 262K server context |
| License? | Apache 2.0 — enterprise-friendly |
| On Hugging Face? | Yes — InternScience/Agents-A1 safetensors |
| Coding? | SciCode 44.3 — competitive in ~35B class, not frontier (GPT-5.5 56.1) |
| vs Qwen AgentWorld? | Different job — AgentWorld simulates envs; Agents-A1 is the acting agent. Shared Qwen-family DNA per HF tags; launch copy doesn't cite AgentWorld |
| 256K enough for agents? | Debatable — beats most open models on long-bench rows, but real agent runs accumulate tool I/O fast. Treat context as necessary, not sufficient |
| How to run? | vLLM or SGLang — not llama.cpp GGUF at ship |
What InternScience claims
Paper title: Scaling the Horizon, Not the Parameters: Reaching Trillion-Parameter Performance with a 35B Agent (arxiv:2606.30616, dated June 29, 2026).
Two scaling axes:
-
Long-horizon trajectories — a domain-grounded knowledge-action infrastructure that jointly constructs actions, observations, and verifier outcomes so multi-step agent runs become trainable targets, not one-off demos.
-
Heterogeneous agent abilities — a three-stage pipeline:
- Full-domain supervised fine-tuning for broad agent behaviors
- Domain teacher models for specialized expertise (search vs science vs engineering)
- Multi-teacher, multi-domain on-policy distillation with heterogeneity-aware optimization
That framing rhymes with Qwen 3.7-Max environment scaling and self-harness agent improvement — the 2026 consensus is that agent quality scales with trajectory diversity, not just parameter count.
InternScience also open-sourced an evaluation framework in the repo (Agents-A1/evaluation) so others can reproduce agent-capability scores under one protocol — a move toward the standardized eval hygiene we cover in the AI benchmarks guide.
Benchmark table — where 35B punches up
Legend from the model card: 🥇 overall SOTA · 🟢 best among ~35B comparables
Long-horizon search
| Benchmark | Agents-A1 | Qwen3.6-35B-A3B | Kimi-K2.6 | GPT-5.5 (xhigh) |
|---|---|---|---|---|
| BrowseComp | 🟢 75.51 | 67.93 | 83.2 | 🥇 84.4 |
| XBench-DS-2510 | 🟢 86.0 | 71.0 | 🥇 90.0 | 84.0 |
| Seal-0 | 🥇 56.36 | 38.74 | 50.45 | 42.34 |
| GAIA | 🟢 96.04 | 78.64 | 80.58 | 87.38 |
explainx.ai read: GAIA 96 and BrowseComp 75.5 are the headline numbers for teams building search + tool agents — the same benchmark family LongCat-2.0 cites at 79.9 (different harness/protocol — always compare apples-to-apples).
Scientific research
| Benchmark | Agents-A1 | Qwen3.6-35B-A3B | DeepSeek-V4-pro |
|---|---|---|---|
| FrontierScience-Olympiad | 🥇 79.0 | 60.3 | 76.0 |
| FrontierScience-Research | 🥇 40.0 | 2.9 | 13.3 |
| HiPhO | 🥇 46.4 | 37.7 | 38.7 |
| HLE w/ tools | 🟢 47.6 | 36.2 | 48.2 |
Strongest story: research-agent tasks with tools — relevant for RAG + calculator + literature search stacks, not pure chat.
Instruction following
| Benchmark | Agents-A1 | Qwen3.6-35B-A3B | GPT-5.5 |
|---|---|---|---|
| IFBench | 🥇 80.61 | 64.4 | 75.9 |
| IFEval | 🥇 94.82 | 91.3 | 93.35 |
| LongBench-v2 | 🟢 60.2 | 57.7 | — |
If your product fails on multi-constraint prompts (format + content + exclusions), this row matters more than MMLU.
Engineering / coding (the skeptical row)
| Benchmark | Agents-A1 | Qwen3.6-35B-A3B | Kimi-K2.6 | GPT-5.5 |
|---|---|---|---|---|
| SciCode | 🟢 44.33 | 35.8 | 53.5 | 🥇 56.1 |
| MLE-Lite | 🟢 43.94 | 34.85 | 62.12 | 🥇 72.73 |
X asked "What about coding?" — fairly. Agents-A1 wins its weight class but Kimi K2.6 and GPT-5.5 lead on SciCode. For repo-scale coding agents, prioritize Kimi K2.7-Code, LongCat-2.0, or dense local Qwen 3.6 27B until independent Terminal-Bench / SWE-bench runs appear.
Architecture and lineage
| Field | Value |
|---|---|
| Parameters | 35.11B (MoE) |
| Format | Safetensors, BF16 |
| Architecture tag | qwen3_5_moe |
| Context | 256K native; servers document 262144 max |
| Modalities | Text + vision encoder (text-only mode skips vision to free KV cache) |
| License | Apache 2.0 |
The qwen3_5_moe tag and ModelScope distribution place Agents-A1 in the Alibaba / Qwen ecosystem orbit — alongside Qwen-AgentWorld (35B-A3B world model, June 23) and the China free-models playbook.
AgentWorld vs Agents-A1:
| Qwen-AgentWorld | Agents-A1 | |
|---|---|---|
| Role | Predict environment observations | Execute agent policy (plan, call tools) |
| Training focus | World-model / simulation RL | Multi-teacher agent distillation |
| Open weights | 35B-A3B MoE | 35B MoE |
X commenters suggesting "basically Qwen AgentWorld rebranded" oversimplify — but teams evaluating both should read AgentWorld's paper and Agents-A1's distillation story as complementary, not duplicate.
How to run Agents-A1 (vLLM and SGLang)
Weights are Transformers safetensors — use vLLM or SGLang, not llama.cpp until community GGUF quantizers ship.
SGLang — standard server
uv venv --python 3.12 --seed --managed-python
source .venv/bin/activate
uv pip install sglang
SGLANG_USE_MODELSCOPE=true python -m sglang.launch_server \
--model-path InternScience/Agents-A1 \
--port 8000 \
--tp-size 1 \
--mem-fraction-static 0.8 \
--context-length 262144 \
--reasoning-parser qwen3
OpenAI-compatible API: http://localhost:8000/v1
SGLang — tool calling
SGLANG_USE_MODELSCOPE=true python -m sglang.launch_server \
--model-path InternScience/Agents-A1 \
--port 8000 \
--tp-size 1 \
--mem-fraction-static 0.8 \
--context-length 262144 \
--reasoning-parser qwen3 \
--tool-call-parser qwen3_coder
vLLM — standard server
uv venv --python 3.12 --seed --managed-python
source .venv/bin/activate
uv pip install vllm --torch-backend=auto
VLLM_USE_MODELSCOPE=true vllm serve InternScience/Agents-A1 \
--port 8000 \
--tensor-parallel-size 1 \
--max-model-len 262144 \
--reasoning-parser qwen3
vLLM — tool calling
VLLM_USE_MODELSCOPE=true vllm serve InternScience/Agents-A1 \
--port 8000 \
--tensor-parallel-size 1 \
--max-model-len 262144 \
--reasoning-parser qwen3 \
--enable-auto-tool-choice \
--tool-call-parser qwen3_coder
Text-only (save KV cache)
VLLM_USE_MODELSCOPE=true vllm serve InternScience/Agents-A1 \
--port 8000 \
--tensor-parallel-size 1 \
--max-model-len 262144 \
--reasoning-parser qwen3 \
--language-model-only
Recommended sampling (from model card)
| Parameter | Value |
|---|---|
temperature | 0.85 |
top_p | 0.95 |
top_k | 20 |
min_p | 0.0 |
presence_penalty | 1.1 |
repetition_penalty | 1.0 |
Wire to agent harnesses
Point any OpenAI-compatible harness at http://localhost:8000/v1:
- OpenCode —
opencode.jsoncbaseURL - Codex OSS — local provider
- Claude Code / OpenClaw — if your stack supports custom OpenAI endpoints
For MCP-heavy loops, run the tool-call parser variant and validate with your real server set — benchmark tool rows don't guarantee clean JSON on your schema.
What X got right (and what to verify)
Reasonable hype
- IFEval / IFBench SOTA — instruction-following is a real product surface; numbers are strong.
- FrontierScience-Research 40.0 — large jump vs 35B peers (2.9–6.7 range in table).
- Open eval code — rare and useful; run it before procurement.
Skepticism to keep
- "256K is not enough for agent" — long-horizon agents blow context with tool payloads, retries, and state snapshots. 256K helps; state externalization and compaction still required.
- "7 steps before state drift" (Gregor's reply) — real-world agent runs fail on memory coherence, not just benchmark max scores.
- Coding — wait for Terminal-Bench 2.0 / SWE-bench reproductions on your harness before replacing Kimi/LongCat/Qwen dense locals.
- Vendor tables — when a model doesn't report a benchmark, InternScience says they evaluated under their protocol; cross-check against original vendor papers.
Where Agents-A1 fits in the 2026 open-agent ladder
Coding-first open MoE → LongCat-2.0, Kimi K2.7-Code
Local daily driver (dense) → Qwen 3.6 27B + llama.cpp
World-model / sim RL → Qwen-AgentWorld
Heterogeneous long-horizon → Agents-A1 ← this launch
Closed frontier (if allowed)→ GPT-5.5, Fable/Mythos (offline)
During the US export ban on Fable/Mythos, Apache 2.0 agent models that excel at search + science + instructions fill a different niche than GLM-5.2 coding plans — tier workloads like explainx.ai recommends in closed vs open alternatives.
Update — June 29, 2026: Apodex-1.0-mini — another Qwen3.5-35B-A3B agent fine-tune — topped FutureX at 59.17, beating Sonnet 4.6 and GPT-5.5 on future prediction. Agents-A1 targets heterogeneous benchmarks; Apodex targets verification-centric deep research.
Reproduction checklist
- Pull weights —
huggingface-cli download InternScience/Agents-A1or ModelScope mirror - Serve with tools — vLLM/SGLang commands above on a GPU box with headroom for 262K KV
- Clone eval —
Agents-A1/evaluationfrom the GitHub repo linked on Hugging Face - Run your agent loop — same MCP/search tools you use in production
- Compare MoE vs dense — if instruction-following regresses on repo tasks, try Qwen 3.6 27B dense locally
- Log horizon length — steps until drift; benchmarks won't show this
Add verification loops from explainx.ai loops when agents drive CI.
Related on explainx.ai
- Qwen-AgentWorld — language world model — environment simulation vs agent policy
- Qwen 3.6 27B local dev — dense vs MoE — hands-on MoE skepticism
- LongCat-2.0 — 1.6T MoE coding agent — same-day launch, coding focus
- Kimi K2.7-Code — open coder alternative
- Qwen 3.7-Max — long-horizon autonomy — Alibaba agent scaling thesis
- Run open-source models in OpenCode — wire local APIs to harnesses
- AI benchmarks complete guide — GAIA, Terminal-Bench, how to read vendor tables
- Top Chinese AI companies 2026 — ModelScope / Qwen ecosystem map
- What are AI agents? — agent loop fundamentals
Official: Hugging Face — InternScience/Agents-A1 · ModelScope model page · Technical report (arxiv:2606.30616)
Benchmark figures and sampling defaults reflect the InternScience model card as of June 30, 2026. MoE serving requirements and independent coding evals may differ — verify before production. Last updated: June 30, 2026.