Ask a question and watch three free language models answer side-by-side, live, in real time.
Compare how Solar Pro 4, Tencent Hy3, and Step 3.7 Flash think about the same prompt —
no sign-up, no API keys on this end, just paste your Nous Portal token and go.
Solar Pro 4 — 524K ctx Tencent Hy3 — 262K ctx Step 3.7 Flash — 262K ctx All free via Nous Portal
Authentication: These models are hosted on the
Nous Portal.
Paste your free API token below to start chatting — or just click Connect
to see where to get one.
Live Chat Arena — type once, all three answer
Type a question below, or pick a suggestion.
Press Enter to send (Shift+Enter for newline).
Press Esc to stop streaming.
Solar Pro 4
524K ctxby Upstage
"A large language model for agentic workflows, office productivity, and document-intensive work. Strong on coding and long-context tasks."
💬
Waiting for the first word…
Solar hasn't said anything yet. Ask something!
Tencent Hy3
262K ctxMoE · 295B
"A 295B-parameter Mixture-of-Experts model (21B active per token, 192 experts, top-8 routing). Built for reasoning and production-grade agentic workflows."
💬
Waiting for the first word…
Hy3 hasn't said anything yet. Ask something!
Step 3.7 Flash
262K ctx+ vision
"A 196B-parameter MoE model with a vision encoder, activating ~11B per token. High-efficiency multimodal model with mandatory Chain-of-Thought reasoning."
💬
Waiting for the first word…
Step Flash hasn't said anything yet. Ask something!
Benchmark Arena — 10 prompts, 3 models, side by side
Ready
Each model gets the same 10 prompts across reasoning, creativity, coding, humor, and factual knowledge.
Answers stream in real time in each column. When all 10 are done, click View Results to compare.
Solar Pro 4
524K ctx
Upstage · agentic workflows, coding, long documents
⏳
Benchmark answers will appear here
Run the benchmarks to start
Tencent Hy3
262K ctx
Tencent · 295B MoE, strong reasoning
⏳
Benchmark answers will appear here
Run the benchmarks to start
Step 3.7 Flash
262K ctx
StepFun · 196B MoE + vision, mandatory reasoning
⏳
Benchmark answers will appear here
Run the benchmarks to start
🔑 Connect to the Nous Portal
The three models on this page are hosted by the
Nous Research
inference API. To use them you need a free API token from the Nous Portal.
Go to nousresearch.com and sign up (or log in)
Find your API token in your account settings
Paste it below and click Connect
Your token stays in your browser only — it's never sent to any server other than the Nous API.
📊 Benchmark Results
10 prompts · 3 models · side-by-side comparison. Hover or scroll inside a cell to read the full answer.