NX
NXplace
March 2026 just delivered three model drops that fundamentally shift the "free tier" landscape for AI developers.
Within a 30-day window, Alibaba released Qwen3.6 Plus Preview (completely free on OpenRouter), Google dropped Gemini 3.1 Flash Lite at aggressive pricing, and the Qwen3.5 Plus family matured into a legitimate production workhorse with open-weight siblings.
If you're building agentic workflows, coding agents, or high-volume inference pipelines — this comparison matters. Not in theory. In your monthly API bill.
Here's everything I pulled from OpenRouter benchmarks, Artificial Analysis data, Qubrid early testing, and Google's official model cards. No marketing spin. Just the numbers.
| Spec | Qwen3.6 Plus Preview | Qwen3.5 Plus | Gemini 3.1 Flash Lite |
|---|---|---|---|
| Released | Mar 30–31, 2026 | Feb 2026 | Mar 3, 2026 |
| Context | 1M tokens | 1M tokens | 1M tokens |
| Max Output | 65,536 tokens | ~32,768 tokens | 65,536 tokens |
| Architecture | Next-gen hybrid (closed) | Hybrid MoE, 397B/17B active | Closed |
| Modalities | Text only | Text + Image + Video | Text + Image |
| Reasoning | Always-on CoT (no toggle) | Toggle: on/off | 3 levels: none/low/high |
| Speed (tok/s) | 45 | ~12 (~39s avg response) | 381.9 |
| Input Price | $0.00/1M | $0.26/1M | $0.25/1M |
| Output Price | $0.00/1M | $1.56/1M | $1.50/1M |
| Source | Closed (data collected) | Open weights available | Closed |
| Status | Preview | GA stable | Developer Preview |
Three models. Three very different strategies. Let's break down what each one actually does.
The biggest story about Qwen3.6 Plus isn't its architecture — it's the overthinking fix.
If you've wrestled with Qwen3.5 Plus, you know the complaints:
"It overthinks simple tasks and burns 30 seconds reasoning for a one-sentence answer."
Qwen3.6 Plus Preview directly addresses this. Early benchmark data from Qubrid shows dramatic improvements:
| Metric | Qwen3.6 Plus | Qwen3.5 Plus | Improvement |
|---|---|---|---|
| Consistency Score | 10.0 | 9.0 | ↑ 11% |
| Flaky Test Count | 0 | 2 | ✅ Eliminated |
| Avg Response Time | ~13.9s | ~39.1s | ↑ 3x faster |
| Reasoning Efficiency | Fewer tokens, better output | Over-expanded chains | More decisive |
The 3.6 Plus architecture is described as "next-generation hybrid" — not a standard MoE. Key design choices:
While Alibaba hasn't published full benchmark tables for this specific preview release yet, third-party data is telling a clear story:
| Benchmark | Qwen3.6 Plus | Competitor |
|---|---|---|
| Terminal-Bench 2.0 | 61.6 | Claude Opus 4.6: 59.3 |
| OmniDocBench v1.5 | 91.2 | Claude Opus 4.6: 87.7 |
| SWE-bench Verified | 72.4 | GPT-5 mini: 72.4 (tie) |
| Claw-Eval (real-world agents) | 58.7 | Claude: 59.6 (essentially tied) |
The OpenRouter dashboard shows real production usage across major coding agents:
Top apps using it: Kilo Code (131B tokens), OpenClaw (104B tokens), Cline (60.4B tokens), Claude Code (42.5B tokens), Hermes Agent (41.2B tokens).
Category Rankings on OpenRouter: Programming #3, Academia #12, SEO #38, Finance #26, Legal #42.
Qwen3.5 Plus is the hosted API equivalent of the open-weight Qwen3.5-397B-A17B model. Where it earns its keep:
enable_thinking: true/false per request| Benchmark | Score | Context |
|---|---|---|
| BFCL-V4 (function calling) | 72.2 | Beats GPT-5 mini's 55.5 by 30% |
| SWE-bench Verified | ~76.4 | Strong but behind commercial leaders |
| MMLU-Pro | 87.8 | Frontier-adjacent range |
| MMMLU (multilingual) | 88.5 | Behind Gemini 3 Pro (90.6) but big jump from Qwen3 (84.4) |
✅ Multimodal: Text + image + video — all three models support 1M context, but only Qwen3.5 Plus processes all modalities in that window
✅ Controllable reasoning: Toggle thinking mode per-request. Hard tasks get deep CoT, easy tasks get fast direct output. This is the architectural sweet spot that Qwen3.6's "always-on" and Gemini's "3-level" both try to replicate differently.
✅ Open-weight escape hatch: If cost, sovereignty, or customization matters, you can self-host Qwen3.5-397B-A17B. The 3.6 Plus Preview doesn't offer this option at all.
✅ GA stability: Not a preview. Has a production track record.
Qwen3.5 Plus is powerful but the average response time of ~39.1 seconds tells the story. The model frequently over-expands its reasoning chains on tasks that don't need it. This is precisely the problem Qwen3.6 Plus Preview was built to solve.
Best use pattern for Qwen3.5 Plus: Route requests by complexity. Turn thinking off for extraction/classification, on for complex reasoning. The toggle is the key architectural advantage — you control the compute budget.
Google's fastest model ever, period. The 381.9 tok/s number isn't just a marketing flex — Artificial Analysis ranks it third globally at that speed, behind only Mercury 2 (768 tok/s) and Granite 3.3 8B (438 tok/s). It's the fastest closed-weight model from any major lab.
| Metric | Value |
|---|---|
| Output Speed | 381.9 tok/s |
| Speed vs Qwen3.6 Plus | 8.5x faster |
| Speed vs Qwen3.5 Plus | 16x faster |
| TTFT vs 2.5 Flash | 2.5x faster |
| Intelligence Index (AA) | 34 (up from 21 for 2.5 Flash) |
| Benchmark | Score |
|---|---|
| GPQA Diamond (PhD-level science) | 86.9% |
| MMMU-Pro (multimodal understanding) | 76.8% |
| Video-MMMU | 84.8% |
| Arena Elo | 1432 |
For context: 86.9% GPQA Diamond puts Flash-Lite ahead of older Gemini models that sat in a higher tier. That's unusual for a "lite" model.
This is the feature nobody's talking about enough. Google baked three reasoning levels directly into the API:
none → Max speed (381 tok/s), minimum costlow → Balanced reasoning for dashboards, form fillinghigh → Full step-by-step analysis for complex reasoningThis collapses your entire model routing stack into a single API. Instead of maintaining two models (cheap fast + expensive smart) with custom routing logic, you get one model with a per-request reasoning budget dial.
I've seen the pattern where teams build custom orchestrators that classify task complexity, then route to different models. Think levels is essentially Google saying: "Just use one model and adjust the knob."
At $0.25/1M input + $1.50/1M output, Flash-Lite is:
For a 1,000-request-per-day workload with ~400 token responses: Flash-Lite costs ~$227/year vs. $372/year for 2.5 Flash. That's $145 saved. At enterprise scale, the savings compound fast.
Let's put real workloads against all three models:
Scenario: 500 coding agent requests/day, avg 3K input + 2K output tokens per request
| Model | Daily Cost | Monthly Cost | Notes |
|---|---|---|---|
| Qwen3.6 Plus Preview | $0.00 | $0.00 | Free during preview |
| Gemini 3.1 Flash Lite | $1.88 | ~$56 | Fast, cheap |
| Qwen3.5 Plus | $2.10 | ~$63 | Slightly more expensive |
The 1M context cost comparison matters too. Claude Opus 4.6 charges $5.00/1M input. A single 1M-token request to Claude costs $5.00 just for input. The same request costs $0.00 on Qwen3.6 Plus or $0.25 on Gemini. That's a 20x–1,000x cost differential for long-context workloads.
Here's the practical answer to "which model for what":
| Scenario | Winner | Why |
|---|---|---|
| Agentic coding agents | 🏆 Qwen3.6 Plus | #3 programming, 0 flaky, always-on CoT, free |
| Real-time / speed-critical | 🏆 Gemini 3.1 Flash | 381 tok/s is untouchable. 2.5x TTFT |
| Multimodal (images/video) | 🏆 Qwen3.5 Plus | Text + image + video. Others are limited |
| Self-hosting / sovereignty | 🏆 Qwen3.5 Plus | Only one with open weights |
| Long-context RAG (1M) | 🏆 Qwen3.6 Plus | Free + 65K output vs 32K for others |
| Production stability | 🏆 Qwen3.5 Plus | Only GA model. Others are preview |
| Controllable reasoning | 🏆 Gemini 3.1 Flash | 3 thinking levels > toggle > always-on |
| Cost-sensitive dev | 🏆 Qwen3.6 Plus | Free. Can't beat free. |
If I were building an agentic system today, this is the stack I'd deploy:
| Layer | Model | Role |
|---|---|---|
| Hard reasoning | Qwen3.6 Plus | Code review, repo analysis, multi-step agents |
| Multimodal tasks | Qwen3.5 Plus | Image understanding, video analysis |
| Fast routing/classification | Gemini 3.1 Flash | 381 tok/s for task classification, moderation |
| Fallback production | Qwen3.5 Plus or Gemini 3.1 Flash | GA stability when 3.6 preview changes |
The thinking-level analogy for Gemini 3.1 Flash is like dynamically adjusting reasoning_effort on every request. Qwen3.6's always-on CoT is more like setting temperature=0.7 permanently — consistent, but you can't tune it down. Qwen3.5 Plus sits in the middle with its toggle.
For agents, the 0 flaky test count on Qwen3.6 Plus matters more than any benchmark. In production, flakiness is the difference between a $50/week API bill and a $500/week API bill from retries.
Bottom line: All three are compelling, but for different architectures. Qwen3.6 Plus is the free, agent-optimized text workhorse. Gemini 3.1 Flash is the speed demon with thinking controls. Qwen3.5 Plus is the multimodal workhorse with an open-weight safety net.
Pick based on what your actual workload pattern looks like — not which benchmark wins gold.